Variable-speed vibration signal feature extraction method and system
By employing equal-angle resampling and intelligent feature extraction methods, combined with order analysis and machine learning algorithms, the accuracy and reliability issues of fault diagnosis for sliding wheel bearings under variable speed conditions were resolved, enabling precise assessment of bearing wear and early fault identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-22
- Publication Date
- 2026-05-15
AI Technical Summary
Under variable speed conditions, traditional hardware-based order analysis methods are complex to install and costly, while computational order analysis techniques are difficult to effectively extract pulley bearing fault characteristics under variable speed conditions, affecting the accuracy and reliability of fault diagnosis.
By employing equal-angle resampling technology combined with intelligent feature extraction and pattern recognition algorithms, and through order analysis, adaptive filtering, and machine learning or deep learning algorithms, a physical model of fault characteristics is constructed to achieve mild, moderate, and severe classification and assessment of faults such as bearing wear.
It improves the accuracy and reliability of bearing fault diagnosis under variable speed conditions, and can accurately capture key features of bearing wear process under complex working conditions, so as to realize early fault warning and health status assessment.
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] This invention relates to the field of signal analysis and processing, and in particular to a method and system for extracting features from variable speed vibration signals. Background Technology
[0002] Currently, pulley fault diagnosis technology under variable speed conditions has received widespread attention. In variable speed operating environments, vibration signals contain richer mechanical state information than during steady-state operation because they contain the system's dynamic response to speed changes. Order analysis technology is considered an effective method for fault diagnosis under variable speed conditions. Existing research has explored the implementation methods and advantages and disadvantages of different order analysis methods, and verified the computational accuracy of these methods. In the development of order analysis, two main categories have emerged: hardware-based order analysis and computational order analysis.
[0003] For fault diagnosis of variable-speed pulley bearings, traditional hardware-based order analysis methods are limited by installation complexity and cost, while computational order analysis techniques (especially bondless methods) have become the mainstream approach due to their high accuracy and low cost. In the future, further optimization of variable-speed signal feature extraction methods by combining deep learning, signal sparse decomposition, and advanced instantaneous parameter estimation algorithms will help improve the accuracy and reliability of pulley bearing fault diagnosis in the Z-device. Summary of the Invention
[0004] To address the aforementioned technical issues, this invention proposes a method and system for extracting features from variable speed vibration signals. It employs equal-angle resampling technology to ensure the accuracy and consistency of signal analysis, and combines intelligent feature extraction and pattern recognition algorithms to achieve mild, moderate, and severe classification and assessment of faults such as bearing wear.
[0005] The first objective of this invention is to provide a method for feature extraction of vibration signals under varying rotational speed conditions, comprising acquiring vibration signal data under varying rotational speed conditions, and further comprising the following steps: Step 1: Preprocess the vibration signal data; Step 2: Use order analysis to extract features from the preprocessed vibration signal data; Step 3: Optimize the feature data using an adaptive filtering algorithm; Step 4: Construct a physical model of the fault characteristics, and perform classification or regression analysis on the characteristic data based on machine learning or deep learning algorithms; Step 5: Input the optimized feature data into the fault feature physical model to perform bearing fault diagnosis.
[0006] Preferably, the data preprocessing includes bandpass filtering, adaptive noise suppression, and data interpolation.
[0007] In any of the above schemes, the preferred method is that the order analysis combines the vibration signal with real-time rotational speed information, and performs resampling and feature extraction with the order as the variable. The formula for the order J is: Where f is the vibration characteristic frequency and f0 is the reference shaft rotation frequency.
[0008] In any of the above schemes, step 4 preferably includes calculating the bearing defect characteristic frequency when not subjected to axial force, and deriving the bearing defect characteristic frequency when subjected to axial force.
[0009] Preferably, in any of the above solutions, the characteristic frequencies of bearing defects when not subjected to axial force include: 1) When the outer ring is fixed and the inner ring rotates with the axis, the rotational frequency f of a single rolling element or cage relative to the outer ring is... Bo for: , , ; 2) When the inner ring is fixed and the outer ring rotates with the axis, the rotational frequency f of a single rolling element or cage relative to the inner ring is... Bi for: , ; 3) The characteristic frequency f when the inner and outer rings of the bearing have defects i for: ; If there are defects in the inner raceway, then The frequency f of the rolling element when it rolls over the defect o for: ; If there are defects in the outer raceway, then The frequency f of the rolling element when it rolls over the defect RS for: ; 4) The characteristic frequency when a single rolling element has a defect is: ; 5) The frequency at which the cage contacts the outer ring is: ; frequency of cage contact with inner ring ; Among them, V B Let l be the velocity of the center B of the rolling element.m D is the pitch circumference of the raceway, d is the diameter of the rolling element, and D m f is the bearing raceway pitch diameter. r V is the rotational frequency of the shaft. A Let D be the velocity at point A where the rolling element contacts the inner raceway. i This is the diameter of the inner raceway.
[0010] Preferably, in any of the above solutions, the bearing defect characteristic frequencies under axial force include: The frequency f of outer ring failure o for: ; The frequency f of inner ring failure i for: ; Frequency f of rolling element failure RS for: ; Cage failure frequency f c for: ; Where α is the angle between the direction of the force on the rolling element and the straight lines of the inner and outer raceways.
[0011] In any of the above schemes, it is preferred that the fault characteristic physical model is used to calculate the fault characteristic frequency by comparing the spectrum of the measured signal with the above four types of fault physical models when a surface damage-type fault occurs in the components of the rolling bearing, so as to obtain an accurate fault location.
[0012] A second objective of this invention is to provide a system for extracting features from vibration signals at varying rotational speeds. This system includes a data acquisition module for acquiring vibration signal data under varying rotational speed conditions, a signal processing module, a feature optimization module, and a pattern recognition module. The signal processing module is used to denoise, normalize, and perform time-frequency analysis on the vibration signal in order to extract feature information; The feature optimization module is used to optimize the feature data using adaptive filtering technology; The pattern recognition module is used to classify or regress feature data based on machine learning or deep learning algorithms. The system employs the method described in the first objective to achieve characteristic recognition of variable speed vibration signals.
[0013] Preferably, the signal processing module uses wavelet transform or empirical mode decomposition for multi-scale feature extraction.
[0014] In any of the above schemes, it is preferred that the pattern recognition module adopts a neural network model.
[0015] This invention proposes a method and system for extracting vibration signal features under varying operating speeds. It uses a high-precision sensor to collect vibration signals of a bearing in real time and combines them with an intelligent signal processing algorithm to extract vibration signal features under multiple operating conditions. Attached Figure Description
[0016] Figure 1 This is a flowchart of a preferred embodiment of the variable speed vibration signal feature extraction method according to the present invention.
[0017] Figure 2 This is a bearing structure diagram of a preferred embodiment of the variable speed vibration signal feature extraction system according to the present invention.
[0018] Figure 3 This is a rolling schematic diagram of an embodiment of the rolling element of the variable speed vibration signal feature extraction system according to the present invention.
[0019] Figure 4 This is a schematic diagram of an embodiment of the variable speed vibration signal feature extraction method according to the present invention, showing the axial force.
[0020] Figure 5 This is a schematic diagram of an embodiment of the constant angle sampling method for feature extraction of variable speed vibration signals according to the present invention.
[0021] Figure 6 This is a schematic diagram of the calculation process of an embodiment of the variable speed vibration signal feature extraction method according to the present invention.
[0022] Figure 7 This is a schematic diagram of a test bench according to an embodiment of the variable speed vibration signal feature extraction device of the present invention.
[0023] Figure 8 This is a schematic diagram of an embodiment of the time-domain signal and frequency-domain transformation of the vibration acceleration of a faulty bearing according to the variable speed vibration signal feature extraction method of the present invention.
[0024] Figure 9 This is a schematic diagram of an embodiment of the computational order analysis of the variable speed vibration signal feature extraction method according to the present invention. Detailed Implementation
[0025] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.
[0026] Example 1 like Figure 1As shown, this embodiment provides a method for extracting features from vibration signals under varying rotational speeds. Step 100 involves acquiring vibration signal data under varying rotational speeds, and the method further includes the following steps: Step 110 is performed to preprocess the vibration signal data, which includes bandpass filtering, adaptive noise suppression, and data interpolation.
[0027] Step 120 involves using order analysis to extract features from the preprocessed vibration signal data. This order analysis combines the vibration signal with real-time rotational speed information, using the order as a variable for resampling and feature extraction. The formula for the order J is: Where f is the vibration characteristic frequency and f0 is the reference shaft rotation frequency.
[0028] Execute step 130, and use an adaptive filtering algorithm to optimize the feature data; Perform step 140 to construct a physical model of the fault characteristics and perform classification or regression analysis on the characteristic data based on machine learning or deep learning algorithms.
[0029] Calculate the characteristic frequency of bearing defects when not subjected to axial force, and derive the characteristic frequency of bearing defects when subjected to axial force.
[0030] The characteristic frequencies of bearing defects when not subjected to axial force include: 1) When the outer ring is fixed and the inner ring rotates with the axis, the rotational frequency f of a single rolling element or cage relative to the outer ring is... Bo for: , , ; 2) When the inner ring is fixed and the outer ring rotates with the axis, the rotational frequency f of a single rolling element or cage relative to the inner ring is... Bi for: , ; 3) The characteristic frequency f when the inner and outer rings of the bearing have defects i for: ; If there are defects in the inner raceway, then The frequency f of the rolling element when it rolls over the defect o for: ; If there are defects in the outer raceway, then The frequency f of the rolling element when it rolls over the defectRS for: ; 4) The characteristic frequency when a single rolling element has a defect is: ; 5) The frequency at which the cage contacts the outer ring is: ; frequency of cage contact with inner ring ; Among them, V B Let l be the velocity of the center B of the rolling element. m D is the pitch circumference of the raceway, d is the diameter of the rolling element, and D m f is the bearing raceway pitch diameter. r V is the rotational frequency of the shaft. A Let D be the velocity at point A where the rolling element contacts the inner raceway. i This is the diameter of the inner raceway.
[0031] The characteristic frequencies of bearing defects under axial force include: The frequency f of outer ring failure o for: ; The frequency f of inner ring failure i for: ; Frequency f of rolling element failure RS for: ; Cage failure frequency f c for: ; Where α is the angle between the direction of the force on the rolling element and the straight lines of the inner and outer raceways.
[0032] The fault characteristic physical model is used to calculate the fault characteristic frequency by comparing the spectrum of the measured signal with the above four types of fault physical models when surface damage occurs in the components of the rolling bearing, thus obtaining accurate fault location.
[0033] Step 150 involves inputting the optimized feature data into the fault feature physical model to perform bearing fault diagnosis.
[0034] Example 2 A method for feature extraction of variable speed vibration signals, the method comprising: Vibration signal data under variable speed conditions are acquired, and data preprocessing techniques are used to denoise and normalize the signals to improve signal quality. The data preprocessing techniques include bandpass filtering, adaptive noise suppression, and data interpolation to improve signal quality and reduce noise interference.
[0035] The vibration signal feature extraction is based on the time-frequency analysis method, including short-time Fourier transform, wavelet transform or empirical mode decomposition, to obtain the energy distribution of different frequency components. The time-frequency analysis method uses short-time Fourier transform and improves the ability to analyze the instantaneous features of the signal by selecting an appropriate window function and window length.
[0036] The wavelet transform employs a multi-scale decomposition method to extract feature information at different time scales, thereby improving the accuracy of signal analysis.
[0037] The empirical mode decomposition method includes ensemble empirical mode decomposition to reduce mode aliasing and improve the reliability of signal decomposition.
[0038] An adaptive filtering algorithm is used to optimize the feature data, reduce the impact of external interference on feature extraction, and improve feature stability. The feature mapping relationship is constructed based on machine learning or deep learning models. Classification or regression models are established by training historical data to achieve effective extraction and identification of vibration signal features under variable speed conditions. The machine learning model includes support vector machine, random forest or K-nearest neighbor algorithm to improve the accuracy of feature classification.
[0039] Example 3 A device for extracting features from vibration signals at varying speeds includes: a data acquisition module for acquiring vibration signal data under varying speed conditions and performing preliminary filtering on the signals.
[0040] The signal processing module is used to denoise, normalize, and perform time-frequency analysis on the vibration signal to extract feature information. The signal processing module uses wavelet transform or empirical mode decomposition to extract multi-scale features.
[0041] The feature optimization module uses adaptive filtering technology to optimize feature data, thereby improving the stability and accuracy of feature extraction.
[0042] The pattern recognition module classifies or regresses feature data based on machine learning or deep learning algorithms to achieve intelligent recognition of vibration signal features. The pattern recognition module adopts a neural network model to improve the robustness and generalization ability of feature classification.
[0043] Example 4 This invention proposes a fault diagnosis and condition monitoring method and device for variable speed pulley bearings. This method comprehensively utilizes intelligent monitoring, signal analysis, and feature extraction technologies to improve the health assessment capability of pulley bearings under complex operating conditions. Firstly, based on the working principle and structural characteristics of rotating equipment, the method studies the key fault mechanisms of variable speed pulley bearings and formulates corresponding monitoring and diagnostic schemes, constructing a complete condition monitoring system. The system uses high-precision sensors to collect bearing vibration signals in real time and combines them with intelligent signal processing algorithms to achieve vibration signal feature extraction under multiple operating conditions.
[0044] In the fault diagnosis process, this invention employs a simulated fault testing method to collect vibration signals under typical fault modes and compare them with signals under normal conditions. This allows for in-depth analysis of signal characteristics and the extraction of key health status indicators. By combining data statistics and machine learning algorithms, a fault classification model is established, and reasonable fault thresholds are set based on extensive experimental data to improve the accuracy and robustness of fault detection.
[0045] Furthermore, the monitoring and diagnostic system proposed in this invention has been validated in actual operating environments. By acquiring operational data from actual equipment, the diagnostic model is continuously optimized, improving the reliability and adaptability of the detection system. The system can be widely applied to the health monitoring of complex mechanical equipment, and is particularly suitable for fault diagnosis of high-speed variable-speed equipment, providing crucial technical support for the safe operation and lifespan prediction of equipment.
[0046] This invention proposes a condition monitoring and fault diagnosis method for variable speed pulley bearings. Based on the research implementation progress, the technology development is divided into three stages: short-term, medium-term and long-term. The complete technical system is gradually promoted from equipment investigation and scheme design to laboratory verification and actual equipment application, so as to ensure the effectiveness and reliability of the monitoring and diagnosis system.
[0047] In the initial research phase, to ensure the scientific validity and applicability of the monitoring and diagnostic system, an in-depth investigation of the rotating equipment's structure, operating conditions, and potential failure modes is necessary. First, key structural parameters of the pulley bearings are obtained through investigation, including bearing type, dimensions, installation method, and load conditions. Simultaneously, operating data of the equipment, such as speed range, working environment, and load variations, are collected to comprehensively understand its working characteristics. Based on this, a monitoring and diagnostic system that meets application requirements is designed, clarifying the system's core functional modules, performance indicators, and implementation methods.
[0048] Subsequently, an experimental design was developed, clarifying the requirements for the experimental setup and preliminary planning of the signal acquisition and processing scheme. This included determining the signal sampling frequency, sampling method (e.g., fixed-interval sampling or equal-angle sampling), and data storage strategy. Appropriate signal processing techniques (e.g., time-frequency analysis, order analysis, instantaneous frequency estimation) were also selected as the foundation for subsequent algorithm research. The completed scheme design will lay a solid theoretical and technical foundation for subsequent experimental research.
[0049] During the mid-term research phase, a combination of simulation experiments and laboratory tests was used to verify the feasibility and effectiveness of the designed scheme. First, computer simulation technology was used to model the operating signals of the variable-speed pulley bearing, simulating the bearing's vibration response under different operating conditions, and an analysis algorithm for this signal was developed. The simulated signals were then subjected to comprehensive analysis in the time domain, frequency domain, and time-frequency domain to extract fault characteristic parameters, thereby verifying the algorithm's accuracy in fault identification.
[0050] After the simulation experiments are completed, actual equipment testing is required in a laboratory environment to further verify the applicability of the signal processing algorithm. Corresponding test benches are built for different types of rotating equipment, and high-precision sensors are deployed to acquire experimental data. Subsequently, various signal processing methods (such as wavelet transform, empirical mode decomposition, and variable speed order analysis) are combined to conduct in-depth analysis of the experimental signals, extracting key fault features and establishing a health status assessment model for the rotating machinery. Finally, a prototype of the condition monitoring and fault diagnosis principle is developed, preparing for the engineering application of the system.
[0051] In the later stages of the research, the monitoring and diagnostic system was applied to actual equipment and tested under real-world operating conditions to verify its reliability and stability in complex environments. First, a monitoring system was built on-site to monitor the operating data of the rotating machinery over a long period, evaluating the system's adaptability to various operating conditions. Based on the measured data, the monitoring and diagnostic algorithm and fault identification thresholds were optimized to improve the system's fault identification accuracy.
[0052] Simultaneously, dynamic experiments covering the entire system were conducted, including actual ship mooring tests, to evaluate the system's operational performance in real-world application environments. To address potential interference factors encountered on-site (such as noise, temperature variations, and equipment aging), the data processing workflow was further optimized to improve the algorithm's anti-interference capabilities. Finally, the final design of the monitoring and diagnostic system was completed, ensuring its adaptability to practical application needs. A comprehensive engineering application analysis report was also compiled, providing effective technical support for intelligent condition monitoring and fault diagnosis of rotating machinery.
[0053] Example 5 To achieve condition monitoring and fault diagnosis of variable speed bearings, this scheme adopts a technical approach combining experimental testing and data analysis. For experimental testing, a variable speed pulley test bench is established, with two vibration acceleration sensors installed on each monitoring pulley, one horizontally and one vertically, and one speed sensor installed on each of the port and starboard sides. These are used to collect the vibration and speed signals of the pulleys. For data analysis, data analysis software is developed, utilizing signal processing methods such as time-domain index monitoring, order analysis, and time-frequency analysis to analyze the vibration signals of the motor, determine the health status of the bearing, and perform condition monitoring and fault diagnosis. Through experimental testing and data analysis, characteristic indicators of bearing wear faults are constructed, and the degree of wear is diagnosed as mild, moderate, or severe, providing maintenance suggestions. The sensor layout and diagnostic methods for variable speed pulley fault diagnosis are shown in Table 1.
[0054] Table 1. Sensor Layout and Diagnostic Methods for Fault Diagnosis of Variable Speed Pulleys Due to the instability of the operating state of pulley bearings under variable speed conditions, the collected vibration signals exhibit highly non-stationary characteristics, showing significant differences from signals under traditional stationary operating conditions. The complexity of this non-stationary signal makes it difficult to effectively extract fault features and accurately identify the bearing's health status using conventional signal analysis methods. Therefore, to overcome this challenge, this study specifically focuses on the wear faults of variable speed pulley bearings, conducting a systematic study on condition monitoring and fault diagnosis. It explores time-domain index construction techniques and order analysis techniques to achieve accurate determination of mild, moderate, and severe faults in pulley bearings. By combining time-domain and frequency-domain analysis, the study comprehensively mines the characteristic information of bearings under variable speed environments, constructs time-domain characteristic indices characterizing the bearing's wear state, and extracts the changing patterns of frequency-domain spectral lines during bearing wear, thereby effectively improving the condition monitoring and fault diagnosis capabilities of variable speed pulley bearings.
[0055] To ensure the accuracy and stability of vibration signal acquisition, vibration acceleration sensors are deployed at each key monitoring point of the pulley system to capture vibration signals generated during bearing operation in real time. If conditions permit, speed sensors can also be installed to synchronously record pulley speed changes, further improving the accuracy of signal analysis. The specific installation locations of the sensors require thorough on-site investigation and optimized layout to ensure that the acquired signals accurately and effectively reflect the operating status of the pulley bearing.
[0056] In terms of fault experiment research, a specialized variable-speed pulley fault test bench was first built to simulate the pulley operating environment under actual working conditions. During the experiment, bearing samples with different degrees of wear were processed, covering normal bearings, lightly worn bearings, moderately worn bearings, and heavily worn bearings. Test experiments were conducted on bearings in different states, and their vibration signals were collected. The vibration characteristics of various bearings under variable-speed operating conditions were compared and analyzed. The effectiveness of the bearing fault feature extraction method was verified through experimental data, and the pattern recognition algorithm was optimized to improve the accuracy of fault diagnosis.
[0057] Because pulley bearings operate under variable speed conditions, traditional Fourier transform methods, limited by the assumption of stationary signals, cannot effectively analyze the spectral characteristics of variable speed signals. Under variable speed conditions, speed changes cause blurring or drifting of speed-related frequency components in the spectrum, making it difficult for traditional frequency domain analysis methods to provide clear and stable feature information. Therefore, this invention employs order analysis technology to extract fault features of pulley bearings. Order analysis combines vibration signals with real-time speed information, using the order as a variable for resampling and feature extraction, thereby effectively eliminating the influence of speed changes on the signal spectrum and enhancing the robustness of feature extraction. This method can accurately capture changes in key order components during bearing wear, ensuring accurate monitoring of the pulley bearing's health status and early fault warning even under complex variable speed conditions.
[0058] Order analysis is one of the most important methods in the vibration analysis of variable-speed rotating machinery, and it is highly effective in processing vibration signals under varying speeds. Order analysis fully utilizes the rotational speed signal, performing equal-angle resampling to transform the variable-speed signal into a stationary signal within the angular domain for signal analysis. Because most discrete frequency components in the vibration signal of rotating machinery are related to the dominant rotational frequency (fundamental frequency), dividing each spectral value on the horizontal axis of the spectrum by a reference value (usually the rotational speed frequency) transforms the horizontal axis into a dimensionless order ratio. Therefore, the order is defined as follows: the order is the number of times the signal oscillates within one revolution of the reference shaft (usually the rotating shaft). Let the vibration characteristic frequency be f, and the reference shaft rotational frequency be f0, then the mathematical expression of the order J is as follows: J=f / f0 In the diagnosis of wear faults in pulley bearings, fault mode identification is also crucial. The wear fault modes of pulley bearings can be effectively identified by analyzing their vibration signal characteristics. When a bearing wears, the time and frequency domain characteristics of its vibration signal undergo significant changes, exhibiting specific fault characteristic components. Based on the fault feature extraction method described above, these characteristic information can be accurately obtained to assist in the accurate identification of fault modes. As the bearing's operating condition gradually deteriorates from mild wear to moderate or even severe wear, the corresponding fault characteristics become more pronounced, and the energy distribution and frequency components of the vibration signal also change systematically. Therefore, this study classifies the wear state of pulley bearings into three levels: mild wear, moderate wear, and severe wear, and establishes classification criteria based on the changing trends of fault characteristics to determine the degree of bearing wear. This fault mode identification method can provide a scientific basis for the health management of pulley bearings, ensuring early detection and timely repair of faults while preventing equipment performance degradation or damage due to fault deterioration.
[0059] To ensure the reliability and effectiveness of the pulley bearing fault diagnosis system in practical applications, parameter thresholds need to be set and optimized based on actual equipment. First, a signal acquisition system is established under real-world operating conditions to ensure stable acquisition of vibration signals during equipment operation and to collect vibration data under normal operating conditions. Then, the acquired signals are analyzed in detail, calculating key time-domain characteristic indicators, including root mean square (RMS), peak value, and kurtosis, to quantify the bearing's health status. Based on the analysis results, the most representative time-domain characteristics are selected as key monitoring indicators, and threshold standards for each level of fault are reasonably set according to the vibration signal characteristic distribution under different wear states. Finally, the rationality and adaptability of the set thresholds are verified through multiple tests under different operating conditions, and the discrimination accuracy of the diagnostic model is further optimized. This experimental research not only improves the accuracy and stability of the fault diagnosis system but also lays a solid foundation for online monitoring and intelligent diagnosis of pulley bearings.
[0060] Rolling bearings consist of four parts: inner ring, outer ring, rolling elements, and cage. Therefore, when surface damage-related failures occur in rolling bearings, the characteristic frequencies of these failures are included in all four types of failure characteristics. Here, physical models of the failure characteristics of the inner ring, outer ring, rolling elements, and cage are established respectively. Specifically... Figure 2 As shown.
[0061] The main geometric parameters of the rolling element bearing shown in the diagram are: Bearing pitch diameter D: The diameter of the circle containing the center of the bearing rolling elements. Rolling element diameter d: the average diameter of the rolling element. Contact angle α: The angle between the direction of the force on the rolling element and the straight lines of the inner and outer raceways. Number of rolling elements Z: The total number of rolling elements.
[0062] In order to accurately analyze the motion parameters of each component of the bearing, the following assumptions are made in this study: (1) There is no relative sliding between the raceway and the rolling elements; (2) Each rolling element has the same diameter and is evenly distributed between the inner and outer raceways; (3) No deformation occurs in any part when subjected to radial and axial loads; Method: First, the characteristic frequencies of bearing defects when not subjected to axial force are studied, and then the characteristic frequencies of bearing defects when subjected to axial force are derived.
[0063] (a) Characteristic frequency of bearing defects when not subjected to axial force i) When the outer ring is fixed and the inner ring rotates with the axis, the rotational frequency of a single rolling element (or cage) relative to the outer ring is shown in the schematic diagram of the rolling element rolling. Figure 3 As shown.
[0064] From figure (a), we can see that the tangential velocity of the inner raceway is: Among them, f r Let d be the rotational frequency of the shaft, and d be the diameter of the rolling element. i D is the diameter of the inner raceway. m This is the bearing raceway pitch diameter, which is the average value of the inner and outer raceways.
[0065] Because the rolling element rolls without slipping, the velocity at point A where the rolling element contacts the inner raceway is: Since the outer ring is fixed, the velocity of the rolling element at the contact point C is: The velocity of the rolling element center B (i.e., the cage velocity) is: The rotational frequency of a single rolling element (or cage) relative to the outer ring is: Among them, l m It is the pitch circle circumference of the raceway.
[0066] ii) When the inner ring is fixed and the outer ring rotates with the axis, the rotational frequency of a single rolling element (or cage) relative to the inner ring is: If the rotation frequency of the outer ring is still f r The tangential velocity of the cage relative to the inner ring then changes from... Figure 4 From (b) we can know that: The rotational frequency of a single rolling element (or cage) relative to the inner ring is: iii) Characteristic frequencies when there are defects in the inner and outer rings of the bearing: If there are defects in the inner raceway, then The frequency at which a rolling element rolls over the defect is: If there are defects in the outer raceway, then The frequency at which a rolling element rolls over the defect is: iv) Characteristic frequency of a defective rolling element: If a single defective rolling element impacts the outer raceway (or outer ring) only once per revolution of its rotation, then its rotational frequency relative to the outer ring is: v) Frequency of rubbing between the cage and the inner and outer rings: The frequency at which the cage contacts the outer ring (equal to the frequency at which the outer ring of a single rolling element passes through): The frequency at which the cage contacts the inner ring (equal to the frequency at which the inner ring of a single rolling element passes through): (b) Characteristic frequency of bearing defects under axial force Because the rolling elements have considerable clearance, when subjected to axial force, the inner and outer rings of the bearing are axially misaligned, and the contact point between the balls and the raceway moves from points A and B to points C and E. For example... Figure 4 As shown.
[0067] At this point, the bearing pitch diameter remains unchanged, but the working diameter of the inner raceway increases and the working diameter of the outer raceway decreases. That is to say, the working diameter of the ball changes from d to dcosα. It is only necessary to replace the formula for calculating the characteristic frequency of bearing defects when there is no axial force (the characteristic frequency of the bearing is only related to the bearing pitch diameter and the ball diameter).
[0068] Based on the above analysis, the physical models for the failure characteristics of the inner ring, outer ring, rolling elements, and cage can be derived as follows: The frequency f of outer ring failure o for: The frequency f of inner ring failure i for: Frequency f of rolling element failure RS for: Cage failure frequency fc for: When surface damage occurs in the components of a rolling bearing, the fault characteristic frequency can be calculated by comparing the spectrum of the measured signal with the physical models of the four types of faults mentioned above, thus obtaining an accurate fault location.
[0069] Order analysis is one of the most important methods in the vibration analysis of variable-speed rotating machinery, and it is highly effective in processing vibration signals under varying speeds. Order analysis technology makes full use of speed signals, effectively analyzing vibration signals under variable-speed conditions, extracting characteristic frequencies of bearing faults, and determining bearing wear faults from the frequency domain based on characteristic data of spectral changes.
[0070] Order analysis is a signal analysis technique developed based on spectrum analysis. Most discrete frequency components of the vibration signal of rotating machinery are related to the dominant rotation frequency (fundamental frequency). By dividing each spectral value on the horizontal axis of the spectrum by a reference value (usually the rotational speed frequency), the horizontal axis becomes a dimensionless order ratio, and the original spectrum becomes an order ratio spectrum. Therefore, the order is defined as follows: the order is the number of times the signal oscillates in one revolution of the reference axis (usually the rotating shaft). Let the characteristic vibration frequency be f, and the reference axis rotation frequency be f0, then the mathematical expression of the order J is as follows: To perform order analysis, data sampled at equal time intervals must be resampled at equal angles to obtain a stationary vibration signal within the angular domain. Under normal conditions, the vibration signals of tested mechanical equipment are sampled at equal time intervals. The sampling is performed using a fixed sampling frequency f. s This equal time interval Sampling is not a problem under constant shaft operating conditions. However, under variable speed operating conditions, due to the non-stationarity of the signal, the data obtained by equal-time sampling does not meet the signal stationarity requirements of the Fourier transform. Therefore, it is necessary to use speed pulses to resample the signal at angles to obtain a stationary vibration signal that satisfies the Fourier transform in the angular domain.
[0071] Using the keyway markings or reflectors on the shaft, only one pulse can be collected per revolution. Equal-angle resampling involves using the collected pulses, each marked with 2π radians per revolution, to interpolate and divide them into N segments using appropriate calculation methods. Each segment has a radian value of... Since the rotational speed pulse cannot reflect whether the shaft is accelerating or moving at a constant speed within the time of one 2π radian revolution, it can be assumed that the shaft is moving with uniform angular acceleration in each small time interval. Figure 5The diagram shows a schematic of equal-angle resampling. Figure 5 In this process, 8 points are sampled per revolution in every two acquired key phase signals. These inserted time points are not at equal time intervals from the time axis perspective, but they are at equal intervals from the angular domain perspective.
[0072] Order analysis techniques utilize rotational speed pulses to resample the signal at angles, obtaining a stationary vibration signal that satisfies the Fourier transform in the angular domain, such as... Figure 6 As shown.
[0073] Compared to traditional order tracking methods, this method does not directly sample the signal synchronously. The key phase information and the vibration signal are acquired and stored simultaneously in two separate streams. During signal processing, the instantaneous rotational speed information of the vibration signal is obtained by calculating the time interval of the key phase signal. This instantaneous rotational speed information is then used in software to obtain a stable vibration signal in the angular domain.
[0074] In preliminary work, the institute analyzed the basic theory and implementation steps of order analysis technology and applied it to bearing fault diagnosis, successfully detecting bearing faults. Simultaneously, to meet project requirements, order analysis was applied to the condition detection and fault diagnosis of pulley bearings in device Z. The institute designed a variable-speed pulley bearing test bench to simulate the working characteristics of pulley bearings and conduct experiments, laying the foundation for practical engineering applications.
[0075] Using an experimental platform, such as Figure 7 As shown, the test bench includes a motor, shaft, frequency converter, and brake. The motor provides power output, the frequency converter controls the motor speed, and the brake simulates the load. The failure characteristics of the bearings are shown in Table 2.
[0076] Table 2. Failure Characteristic Orders of Bearings Adjust the frequency converter to make the input motor speed a deceleration process, with a speed range of 0 r / min to 3000 r / min. Collect vibration signals and key phase signals from the input shaft in the time domain.
[0077] Based on the feature order, the maximum analysis order is set to 50, meaning that 100 points are sampled at equal angles within each revolution of the axis. The sampling angle for equal angle sampling is... Based on the operating conditions and the inverter's speed adjustment range under experimental conditions, the maximum operating speed of this gearbox is set to 3000 r / min. (From the formula...) The minimum sampling frequency in the time domain is f. s=5000Hz. The actual sampling frequency was set to 10240Hz, which meets the sampling frequency setting requirements. The time domain frequency domain analysis of the vibration acceleration signal of the faulty bearing obtained above is shown in Figure 8.
[0078] from Figure 8 As can be seen, the vibration amplitude gradually decreases from 1g to 0g, indicating that the gearbox is decelerating. Spectral analysis of this signal, as mentioned earlier, shows that the frequency components on the frequency axis are not well discretized. Frequency confusion occurs because the fundamental frequency component is continuous, and no obvious spectral lines or sidebands reflecting fault characteristics appear.
[0079] The signal sampled at equal time is resampled at equal angles using a key phase pulse signal, such as... Figure 9 The spectrum diagram shows the order of the bearing outer race fault. A clear peak related to the bearing outer race fault (order 3.048) can be seen near the third harmonic, as well as a similar peak at order 3.125 in the figure.
[0080] Based on the aforementioned gear failure mechanism and the form of failure characteristics, it can be determined that the failure is due to a bearing outer ring failure. Comparing the signal spectrum and order spectrum, the characteristic order spectral peaks representing bearing damage information are prominent and very clear in the order spectrum, and are clearly distinguishable from the surrounding spectral lines, making it easier to determine the bearing condition information under variable speed conditions.
[0081] To better understand this invention, specific embodiments have been described in detail above, but these are not intended to limit the invention. Any simple modifications made to the above embodiments based on the technical essence of this invention still fall within the scope of this invention. Each embodiment in this specification focuses on its differences from other embodiments; similar or identical parts between embodiments can be referred to mutually. For system embodiments, since they basically correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
Claims
1. A method for extracting features from vibration signals under varying rotational speed conditions, comprising acquiring vibration signal data under varying rotational speed conditions, characterized in that, It also includes the following steps: Step 1: Preprocess the vibration signal data; Step 2: Use order analysis to extract features from the preprocessed vibration signal data; Step 3: Optimize the feature data using an adaptive filtering algorithm; Step 4: Construct a physical model of the fault characteristics, and perform classification or regression analysis on the characteristic data based on machine learning or deep learning algorithms; Step 5: Input the optimized feature data into the fault feature physical model to perform bearing fault diagnosis.
2. The method for extracting features from variable speed vibration signals as described in claim 1, characterized in that, The data preprocessing includes bandpass filtering, adaptive noise suppression, and data interpolation.
3. The method for extracting features from variable speed vibration signals as described in claim 2, characterized in that, The order analysis combines the vibration signal with real-time rotational speed information, using the order as a variable for resampling and feature extraction. The formula for the order J is: Where f is the vibration characteristic frequency and f0 is the reference shaft rotation frequency.
4. The method for extracting features from variable speed vibration signals as described in claim 3, characterized in that, Step 4 includes calculating the bearing defect characteristic frequency when not subjected to axial force, and deriving the bearing defect characteristic frequency when subjected to axial force.
5. The method for extracting features from variable speed vibration signals as described in claim 4, characterized in that, The characteristic frequencies of bearing defects when not subjected to axial force include: 1) When the outer ring is fixed and the inner ring rotates with the axis, the rotational frequency f of a single rolling element or cage relative to the outer ring is... Bo for: ， ， ； 2) When the inner ring is fixed and the outer ring rotates with the axis, the rotational frequency f of a single rolling element or cage relative to the inner ring is... Bi for: ， ； 3) The characteristic frequency f when the inner and outer rings of the bearing have defects i for: ； If there are defects in the inner raceway, then The frequency f of the rolling element when it rolls over the defect o for: ； If there are defects in the outer raceway, then The frequency f of the rolling element when it rolls over the defect RS for: ； 4) The characteristic frequency when a single rolling element has a defect is: ； 5) The frequency at which the cage contacts the outer ring is: ； frequency of cage contact with inner ring ； Among them, V B Let l be the velocity of the center B of the rolling element. m D is the pitch circumference of the raceway, d is the diameter of the rolling element, and D m f is the bearing raceway pitch diameter. r V is the rotational frequency of the shaft. A Let D be the velocity at point A where the rolling element contacts the inner raceway. i This is the diameter of the inner raceway.
6. The method for extracting features from variable speed vibration signals as described in claim 5, characterized in that, The characteristic frequencies of bearing defects under axial force include: The frequency f of outer ring failure o for: ； The frequency f of inner ring failure i for: ； Frequency f of rolling element failure RS for: ； Cage failure frequency f c for: ； Where α is the angle between the direction of the force on the rolling element and the straight lines of the inner and outer raceways.
7. The method for extracting features from variable speed vibration signals as described in claim 6, characterized in that, The physical model of fault characteristics is used to calculate the fault characteristic frequency by comparing the spectrum of the measured signal with the physical models of the four types of faults mentioned above when surface damage occurs in the components of rolling bearings, thus obtaining accurate fault location.
8. A device for extracting features from vibration signals at varying speeds, comprising a data acquisition module for acquiring vibration signal data under varying speed conditions, characterized in that, It also includes a signal processing module, a feature optimization module, and a pattern recognition module. The signal processing module is used to denoise, normalize, and perform time-frequency analysis on the vibration signal in order to extract feature information; The feature optimization module is used to optimize the feature data using adaptive filtering technology; The pattern recognition module is used to classify or regress feature data based on machine learning or deep learning algorithms. The system uses the method described in claim 1 to achieve characteristic recognition of variable speed vibration signals.
9. The variable speed vibration signal feature extraction device as described in claim 7, characterized in that, The signal processing module uses wavelet transform or empirical mode decomposition to extract multi-scale features.
10. The variable speed vibration signal feature extraction device as described in claim 7, characterized in that, The pattern recognition module uses a neural network model.