Fault diagnosis method based on rotor dynamic characteristics analysis of blade tip clearance signal
By deeply processing and inversion of the blade tip gap signal, the rotor dynamic characteristics are extracted and the fault diagnosis model is constructed, which solves the problem of difficult measurement of blade disk parameters in the existing technology, and achieves accurate diagnosis and diagnostic accuracy of rotor faults.
Patent Information
- Application Number
- CN202210602759.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-30
- Publication Date
- 2025-01-03
- Estimated Expiration
- 2042-05-30
AI Technical Summary
The prior art is difficult to accurately diagnose the fault of rotating mechanical rotors, especially the difficulty in directly measuring the changes in the blade parameters, resulting in limited accuracy of rotor dynamic characteristics analysis and fault diagnosis.
Using the deep processing method based on the tip gap signal, the typical and general characteristics of the rotor dynamic characteristics are extracted, and the fault diagnosis machine learning model is constructed. The signals are obtained through the tip gap sensor, and the deep processing and inversion are carried out to identify the typical rotor faults. and diagnosis.
It realizes accurate acquisition of rotor dynamic characteristics parameters and accurate diagnosis of faults, improves the fault detection accuracy of rotating machinery, and provides a non-contact and low-intervention measurement method without affecting the operating status of the machinery.
Smart Images
Figure CN115324662B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of health management and intelligent operation and maintenance of rotating machinery, and in particular relates to a fault diagnosis method based on rotor dynamics characteristic analysis of blade tip clearance signals. Background Art
[0002] Rotating machinery such as aircraft engines, steam turbines, and gas turbines are very important high-end equipment. Modern rotating machinery is developing towards high speed, high thrust-to-weight ratio, and high pressure ratio. The rotor system, as its core component, works under extremely harsh conditions such as high temperature, high speed, high dynamic strain, and complex media, which directly determines the performance of rotating machinery. According to statistics, rotor failure accounts for more than 70% of rotating machinery failures. The rotor dynamic characteristic parameters are a direct representation of the operating state of the rotor system, and the monitoring and analysis of the rotor dynamic characteristic parameters are the key basis for guiding the optimal design of the rotor system structure, and are also an important part of the intelligent operation and maintenance and digital twin of rotating machinery.
[0003] Rotating machinery rotor dynamic characteristics monitoring and fault diagnosis usually use vibration and displacement sensors installed on the engine casing or shaft system to measure parameters such as rotor bearing force, vibration and rotor axis trajectory, and based on the parameter change characteristics, realize the dynamic characteristics analysis of rotor vortex (axis trajectory), bending / torsion / axial vibration, bearing force, and the diagnosis of typical faults such as rotor imbalance, misalignment, rotor-static friction, axial movement, and blade loss. The blade disk is the core working component of the rotor system, with a large moment of inertia. Under various typical fault excitations, the parameters (vibration, displacement, etc.) it exhibits change most dramatically, and can best characterize the operating state of the rotor system. However, due to the limitations of the installation position and sensing principle, it is difficult for traditional vibration and displacement sensors to directly measure blade disk parameters (vibration, displacement, etc.), resulting in a significant reduction in the accuracy of the current rotating machinery rotor dynamic characteristics analysis and fault diagnosis.
[0004] In recent years, the online monitoring technology of the status parameters of the core components of the rotor based on blade tip sensing has become a research hotspot in the testing technology of rotating machinery due to its advantages of low intervention, non-contact and direct acquisition. Among them, the blade tip clearance measurement technology uses a non-contact sensor probe (e.g., capacitor, optical fiber, eddy current, etc.) installed on the casing to realize the distance from the blade tip to the sensor end face when all blades of the same level pass through the sensor. It is becoming a standard for testing aircraft engines, steam turbines, etc. Studies have shown that the blade tip clearance parameters have a direct intrinsic relationship with the aerodynamic performance and rotor failure of rotating machinery. If the advantages of the blade tip clearance non-contact online measurement can be used to analyze the dynamic characteristic parameters of the rotor system, it will further guide the design and improvement of the rotor structure, which has outstanding scientific value and broad application prospects. According to the measurement principle, the tip clearance can directly reflect the real-time radial displacement parameters of the blade disk, but the analysis of complex rotor bending vibration, torsional vibration and even bending-torsion combined vibration depends more on the circumferential displacement parameters of multi-stage blade disks. Currently, tip timing measurement is an effective means to obtain the circumferential displacement of the blade disk. How to extract high-precision tip timing information from the lower bandwidth tip clearance signal and further explore the deep-level data features is the key to using it for rotor dynamics characteristic analysis and then realizing fault diagnosis. Summary of the invention
[0005] The purpose of the present invention is to overcome the shortcomings of existing rotor fault diagnosis methods and propose a fault diagnosis method based on the analysis of rotor dynamic characteristics of blade tip clearance signals. The blade tip clearance signal is deeply processed to extract typical characteristics of rotor dynamic characteristics such as rotor vortex (axis trajectory), bending vibration, torsional vibration, axial vibration, and invert the rotor dynamic characteristics. The universal feature information of the blade tip clearance signal is introduced to construct a rotor fault diagnosis machine learning model that integrates the typical features of rotor dynamic characteristics and the universal features of the blade tip clearance signal, so as to achieve accurate identification and diagnosis of typical rotor faults.
[0006] The objective of the present invention is achieved through the following technical solutions:
[0007] A fault diagnosis method based on rotor dynamic characteristics analysis of blade tip clearance signals, using blade tip clearance signals measured by a plurality of blade tip clearance sensors to obtain rotor dynamic characteristics and realize typical rotor fault diagnosis; wherein the blade tip clearance sensor is connected to a signal conditioning module via a connecting line, the signal conditioning module is connected to a signal acquisition module via a connecting line, and the signal acquisition module is connected to a computer via a connecting line; comprising:
[0008] 1) Install several blade tip clearance sensors on the rotating mechanical casing, install at least two blade tip clearance sensors on each blade disc circumferentially, and install blade tip clearance sensors on at least two blade discs;
[0009] 2) When the rotating machinery is running, the computer obtains the blade tip clearance signal; the blade tip clearance signal is deeply processed to obtain the blade number, profile error, blade tip timing and axial clearance;
[0010] 3) extracting typical features of rotor dynamics characteristics and inverting rotor dynamics characteristics according to the blade tip clearance signal, blade tip timing and axial clearance results obtained in step 2);
[0011] 4) according to the blade tip clearance signal obtained in step 2), obtaining the time domain, frequency domain, and time-frequency domain features of the blade tip clearance signal, and extracting the general features of the blade tip clearance signal;
[0012] 5) According to the typical features of the rotor dynamic characteristics extracted in step 3), directly identifying the rotor imbalance fault, axial movement fault and blade missing fault;
[0013] 6) Based on the rotor dynamic characteristics inverted in step 3) and the universal characteristics of the tip clearance signal extracted in step 4), a shallow machine learning model is used to diagnose typical rotor faults.
[0014] Furthermore, the tip clearance sensor is composed of one of a capacitive sensor, an optical fiber sensor, an eddy current sensor, and a microwave sensor, and is installed on a rotating mechanical casing. Several tip clearance sensors are arranged at different axial levels and circumferential angles, and the signal obtained when the blade passes through the tip clearance sensor is a tip clearance signal.
[0015] Furthermore, the rotor dynamic characteristics include four types: rotor vortex, bending vibration, torsional vibration, and axial vibration. The typical features of the rotor dynamic characteristics include the shape and size of the rotor vortex, and also include the amplitude, frequency and phase of the bending vibration, torsional vibration and axial vibration; the rotor vortex characteristics at the blade disk are inverted using the tip clearance signal obtained by the tip clearance sensor at different positions of the same level; the bending vibration characteristics are inverted using the tip clearance signal of the blades with the same number at different levels; the torsional vibration characteristics are inverted using the tip timing of the blades with the same number at different levels; and the axial vibration characteristics are inverted using the axial clearance.
[0016] Furthermore, the universal characteristics of the tip clearance signal include various statistics, correlation functions, probability density function time domain characteristics, spectrum, power spectrum, inverse spectrum, envelope spectrum, coherence spectrum frequency domain characteristics, short-time Fourier spectrum, wavelet spectrum, Hilbert spectrum, spectrogram, scalogram, and Wigner-Ville distribution time and frequency domain characteristics.
[0017] Furthermore, the typical rotor faults include five types of faults: rotor imbalance, rotor misalignment, rotor-static friction, axial movement, and blade missing. The rotor vortex and bending vibration characteristics are used to diagnose rotor imbalance and rotor misalignment faults; the bending vibration and torsional vibration characteristics are used to diagnose rotor-static friction and blade missing faults; and the axial vibration characteristics are used to detect axial movement faults.
[0018] Furthermore, the deep processing method in step 2) includes: using the blade angle fingerprint map to obtain the blade number and contour error; using the centroid method to obtain the blade tip timing; and using the frequency spectrum line relationship to obtain the axial clearance.
[0019] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:
[0020] 1. The tip clearance signal is obtained using a tip clearance sensor, which is installed on the casing of the rotating machine. When the blade sweeps past the tip clearance sensor, the sensed signal is the tip clearance signal. The tip clearance sensor used in the present invention does not contact the rotating blades, and the tip clearance signal measurement is a non-contact, low-intrusion measurement method that will not affect the operating state of the rotating machine.
[0021] 2. The rotor of the rotating machinery can be simplified into a simply supported beam supported at both ends, with the blade disk in the middle of the simply supported beam. The rotor vibration response is the largest at the blade disk position. In addition, the blade disk has a large radius and mass, resulting in a large moment of inertia. The method for diagnosing rotor faults based on the blade tip clearance signal proposed in the present invention can obtain the shaft system vibration and axis trajectory response at the blade disk position of the rotor component with the largest moment of inertia, directly characterize the rotor operating state, and is beneficial to the inversion of rotor dynamic characteristics and typical fault diagnosis.
[0022] 3. The present invention gives full play to the bridge role of rotor dynamic characteristics. By extracting typical characteristics of rotor dynamic characteristics such as the shape and size of rotor vortex (axial trajectory), amplitude, frequency and phase of bending vibration, torsional vibration and axial vibration, the four rotor dynamic characteristics of rotor vortex (axial trajectory), bending vibration, torsional vibration and axial vibration are inverted. Based on the rotor dynamic characteristics, typical faults such as rotor imbalance, axial movement and blade missing can be directly identified.
[0023] 4. It is difficult to accurately diagnose a variety of faults only through the typical characteristics of rotor dynamics. The present invention combines the typical characteristics of rotor dynamics and the universal characteristics of blade tip clearance signals to provide a multi-category and multi-data comprehensive feature space for fault diagnosis, thus achieving accurate diagnosis of typical faults such as rotor imbalance, rotor misalignment, and rotor-static friction. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 It is a block diagram of the fault diagnosis method based on the rotor dynamic characteristics analysis of the blade tip clearance signal.
[0025] Figure 2 This is a block diagram of a typical rotor fault diagnosis method based on rotor dynamic characteristics.
[0026] Figure 3 It is a schematic diagram of the tip clearance sensor layout.
[0027] Figure 4a and Figure 4b This is the schematic diagram of the tip clearance measurement.
[0028] Reference numerals: 1-rotating shaft, 2-blade disk, 3-blade, 4-blade tip clearance sensor, 5-rotating machine casing, 6-blade tip clearance, 7-blade tip clearance signal DETAILED DESCRIPTION
[0029] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0030] The present invention is described in detail below with reference to the accompanying drawings and embodiments. The specific implementation scheme of the fault diagnosis method based on the rotor dynamic characteristics analysis of the blade tip clearance signal includes six parts: blade tip clearance signal measurement, blade tip clearance signal deep processing, rotor dynamic characteristics inversion, blade tip clearance signal pan-feature extraction, typical fault direct identification, and machine learning fault diagnosis. Figure 1 shown.
[0031] This embodiment uses rotor dynamic characteristics as a bridge, analyzes rotor dynamic characteristics based on tip clearance signal measurement and deep processing results to achieve typical rotor fault diagnosis, such as Figure 2 The specific implementation plan is as follows:
[0032] (1) In a rotating machine, the blade disks are generally multi-stage (with multiple blade disks in the axial direction), such as the blade disks of a civil aircraft engine generally have about 20 stages. Multi-stage blade disks do not require the installation of blade tip clearance sensors at all stages, but at least two stages are required. In this embodiment, a capacitive blade tip clearance sensor is installed on the casing of a three-stage rotating machine. The three-stage rotating machine includes a rotating shaft 1, three blade disks 2, and a plurality of blades 3. The blade tip clearance sensors are arranged in a horizontal and vertical direction, with a total of two sensors per stage, such as Figure 3 The distance between the inner wall of the rotary machine casing 5 and the top of the blade 3 is the blade tip clearance 6.
[0033] (2) Using the capacitive blade tip clearance sensor 4 installed on the rotating mechanical casing 5, the blade tip clearance signal 7 is obtained, as shown in FIG. 4a and FIG. Figure 4b As shown, the blade tip clearance signal 7 is collected using FPGA, and the collected blade tip clearance signal is uploaded to the computer through the TCP protocol.
[0034] (3) Perform deep processing on the tip clearance signal. Use the double-edge time identification method to obtain the tip timing: Assume that the maximum value of the tip clearance signal is A, use a constant level of 0.5A to cut the tip clearance signal, and obtain the time t corresponding to the rising and falling edges of the tip clearance signal. 1 and t 2 , calculate the blade tip timing as t = (t 1 +t 2 ) / 2. Calculate the blade angle fingerprint based on the blade tip timing [1] The blade reference fingerprint is obtained by measuring the angle between adjacent blades in static state with an inclinometer, and the blade angle fingerprint is correlated and matched with the blade reference fingerprint. The blade number is obtained by the maximum value of the correlation matching result. [1] The initial tip clearance sequence of each blade {d 1 ,d 2 ,…,d n}, n is the number of blades, calculate the average initial tip clearance d 0 =(d 1 +d 2 +…+d n ) / n, and obtain the contour error sequence {c 1 ,c 2 ,…,c n}, where c i = d i –d 0 (i=1,2,…,n). Using the obtained blade numbers, we can get the measured tip clearance value sequence {d 1 ',d 2 ',…,d n '}, the measured tip clearance sequence {d 1 ',d 2 ',…,d n '} minus the contour error sequence {c 1 ,c 2 ,…,c n}, eliminate the profile errors of each blade. The axial clearance is extracted by mapping the spectrum line ratio of the tip clearance signal to the axial clearance. [2] .
[0035] (4) Using the blade tip clearance signal, blade tip timing, axial clearance and other signals obtained in step (2) and step (3), invert the rotor dynamic characteristics: using the blade tip clearance signals of sensors at different positions of the same level, calculate the distance function between the center of gravity position of the axis trajectory and each point of the axis trajectory, perform Fourier transform, obtain the Fourier descriptor of the axis trajectory, and extract the rotor vortex (axis trajectory) characteristics at the blade disk in combination with the Fourier descriptor [3]The tip clearance of blades with the same number at different levels is used to calculate the shaft deflection y(t), and Fourier transform is performed to obtain the bending vibration amplitude, frequency and phase, thereby realizing the inversion of bending vibration characteristics. 1 ,t 2 ,…t m , m is the number of rotor stages, in this embodiment m=3, combined with the rotor rotation angular velocity ω, the angle difference of the blades with the same number at different stages is calculated as Δθ=ω(t i -t j ), (i, j = 1, 2, ..., m), to achieve torsional vibration characteristic inversion. Perform Fourier transform on the axial gap z(t) to obtain the axial vibration amplitude, frequency and phase, and achieve axial vibration characteristic inversion.
[0036] (5) Using the blade tip clearance signal obtained in step (2), extract various time domain features of the blade tip clearance signal, such as various statistics, correlation functions, and probability density functions; frequency domain features such as signal spectrum, power spectrum, inverse spectrum, envelope spectrum, and coherence spectrum; and time and frequency domain features such as short-time Fourier spectrum, wavelet spectrum, Hilbert spectrum, spectrogram, scalogram, and Wigner–Ville distribution, to achieve universal feature extraction of the blade tip clearance signal.
[0037] (6) Using the rotor dynamics characteristics of rotor vortex (axis center trajectory), bending vibration, torsional vibration, and axial vibration obtained by inversion in step (4), direct identification of typical rotor faults is achieved. Holographic spectrum images of each frequency doubling of rotor vortex (axis center trajectory) are obtained through holographic spectrum analysis. [4] When the holographic spectrum image is a circle or an ellipse, the rotor imbalance fault is identified. By comparing the amplitude, frequency, and phase characteristics of the axial vibration with the fault threshold, when the vibration amplitude is greater than the fault threshold, the rotor axial movement fault is identified. By comparing the shape of the rotor vortex (axis center trajectory) and the amplitude, frequency, and phase characteristics of the bending vibration with the fault threshold, when the vibration amplitude is greater than the fault threshold, the rotor blade missing fault is identified.
[0038] (7) Using the typical features of the rotor dynamics characteristics extracted in step (4) and the general features of the blade tip clearance signal extracted in step (5), principal component analysis is used to perform feature dimensionality reduction [5] , based on the support vector machine classification model [5] , realize the classification of rotor imbalance, rotor misalignment, rotor-static rubbing, and diagnose typical faults such as rotor imbalance, rotor misalignment, rotor-static rubbing, etc.
[0039] Finally, it should be pointed out that the above examples are only used to illustrate the calculation process of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above examples, a person skilled in the art should understand that the calculation process described in the above examples can still be modified, or some parameters can be replaced by equivalents, and these modifications or replacements do not make the essence of the corresponding calculation method deviate from the spirit and scope of the calculation method of the present invention.
[0040] The present invention is not limited to the embodiments described above. The above description of the specific embodiments is intended to describe and illustrate the technical solution of the present invention. The above specific embodiments are merely illustrative and not restrictive. Without departing from the scope of the present invention and the scope of protection of the claims, a person of ordinary skill in the art can also make many forms of specific changes under the guidance of the present invention, which all fall within the scope of protection of the present invention.
[0041] References:
[0042] [1] Jiang Jiajia, Shao Xingchen, Duan Fajie, et al. A blade automatic positioning method without speed synchronization sensor:, CN110470382A[P].2019.
[0043] [2] Duan Fajie, Li Fafu, Niu Guangyue, et al. Radial and axial clearance measurement system and method for low-pressure turbine crowned blade grate teeth: CN202210238665.0[P].2022.
[0044] [3] Fu Bo, Zhou Jianzhong, Chen Wenqing, et al. An automatic identification method of axis trajectory based on Fourier descriptor [J]. Automation of Electric Power Systems, 2004(12):40-44.
[0045] [4] Qu Liangsheng, Liu Xiong. Principle and application of holographic spectrum analysis method [J]. Equipment Management and Maintenance, 1991(8):3.
[0046] [5]Deng Z, Quan C, Duan FA robust fault diagnosis approach for large-scale production process[J]. Measurement, 2020,170:108737.
Claims
1. A fault diagnosis method based on rotor dynamics characteristic analysis of blade tip clearance signal, characterized in that: The blade tip clearance signals measured by several blade tip clearance sensors are used to obtain rotor dynamic characteristics and realize typical rotor fault diagnosis; wherein the blade tip clearance sensor is connected to the signal conditioning module through a connecting line, the signal conditioning module is connected to the signal acquisition module through a connecting line, and the signal acquisition module is connected to the computer through a connecting line; including: 1) Install several blade tip clearance sensors on the rotating mechanical casing, install at least two blade tip clearance sensors on each blade disc circumferentially, and install blade tip clearance sensors on at least two blade discs; 2) When the rotating machinery is running, the computer obtains the blade tip clearance signal; the blade tip clearance signal is deeply processed to obtain the blade number, profile error, blade tip timing and axial clearance; 3) According to the tip clearance signal, tip timing and axial clearance results obtained in step 2), typical characteristics of rotor dynamics are extracted to invert rotor dynamics; the rotor dynamics include four types: rotor vortex, bending vibration, torsional vibration and axial vibration. The typical characteristics of rotor dynamics include the shape and size of rotor vortex, and also include the amplitude, frequency and phase of bending vibration, torsional vibration and axial vibration; the tip clearance signal obtained by the tip clearance sensor at different positions of the same level is used to invert the rotor vortex characteristics at the blade disk; the tip clearance signal of the blades with the same number at different levels is used to invert the bending vibration characteristics; the tip timing of the blades with the same number at different levels is used to invert the torsional vibration characteristics; the axial clearance is used to invert the axial vibration characteristics; 4) According to the blade tip clearance signal obtained in step 2), the time domain, frequency domain, and time-frequency domain characteristics of the blade tip clearance signal are obtained, and the universal characteristics of the blade tip clearance signal are extracted; the universal characteristics of the blade tip clearance signal include various statistics, correlation functions, and probability density function time domain characteristics, spectrum, power spectrum, inverse spectrum, envelope spectrum, coherence spectrum frequency domain characteristics, short-time Fourier spectrum, wavelet spectrum, Hilbert spectrum, spectrogram, scalogram, and Wigner-Ville distribution time and frequency domain characteristics; 5) According to the typical features of the rotor dynamic characteristics extracted in step 3), directly identifying the rotor imbalance fault, axial movement fault and blade missing fault; 6) According to the rotor dynamic characteristics inverted in step 3) and the universal characteristics of the blade tip clearance signal extracted in step 4), a shallow machine learning model is used to diagnose typical rotor faults; the typical rotor faults include five types of faults: rotor imbalance, rotor misalignment, rotor-static rubbing, axial movement, and blade missing; rotor vortex and bending vibration characteristics are used to diagnose rotor imbalance and rotor misalignment faults; bending vibration and torsional vibration characteristics are used to diagnose rotor-static rubbing and blade missing faults; and axial vibration characteristics are used to detect axial movement faults.
2. The fault diagnosis method based on rotor dynamic characteristics analysis of blade tip clearance signal according to claim 1 is characterized in that: The tip clearance sensor is composed of one of a capacitive sensor, an optical fiber sensor, an eddy current sensor, and a microwave sensor, and is installed on a rotating mechanical casing. Several tip clearance sensors are arranged at different axial levels and circumferential angles. The signal obtained when the blade passes through the tip clearance sensor is a tip clearance signal.
3. The fault diagnosis method based on rotor dynamic characteristics analysis of blade tip clearance signal according to claim 1 is characterized in that: The deep processing method described in step 2) includes: using the blade angle fingerprint map to obtain the blade number and contour error; using the centroid method to obtain the blade tip timing; and using the frequency spectrum line relationship to obtain the axial clearance.