A method for monitoring torsional vibration of a rotating shaft system based on an optical fiber sensor

Abstract

The application discloses a rotating shaft system torsional vibration monitoring method based on an optical fiber sensor and belongs to the technical field of fault diagnosis. The method comprises the following steps: obtaining a torsional vibration signal based on an optical fiber sensor, obtaining a torsional vibration real-time frequency by using a torsional vibration signal frequency estimation method of ASA-LSF, obtaining a torsional vibration natural frequency of a rotating shaft system by using modal analysis of ANSYS, and comparing the torsional vibration real-time frequency with the torsional vibration natural frequency of the shaft system. If a difference between the torsional vibration real-time frequency and the torsional vibration natural frequency of the shaft system is within a set threshold value, it is considered that torsional vibration resonance is prone to occur at this time, and a resonance early warning is issued. If the difference between the torsional vibration real-time frequency and the torsional vibration natural frequency of the shaft system is beyond the set threshold value, it is considered that torsional vibration resonance is not prone to occur at this time, and no resonance early warning is issued. The application can get rid of the dependence on prior knowledge such as signal processing technology and diagnosis experience, and has the characteristics of simplicity, directness and convenience.

Description

A method for monitoring torsional vibration of rotating shaft systems based on fiber optic sensors Technical Field

[0001] This invention belongs to the field of fault diagnosis technology and relates to the diagnosis of torsional vibration resonance in rotating shaft systems. Specifically, it is a method for monitoring torsional vibration in rotating shaft systems based on fiber optic sensors. Background Technology

[0002] Rotating machinery such as electric motors, internal combustion engines, and steam turbines are the most widely used mechanical equipment, extensively applied in fields such as power, energy, chemical industry, and shipbuilding. Their safe operation directly affects the stability of the entire production process. Torsional vibration is one of the main vibration forms of rotating shaft systems. Torsional vibration is caused by an imbalance in the total torque resulting from the action of external periodic excitation torque, resulting in alternating motion or corresponding deformation along the circumferential direction of the rotating shaft axis. Therefore, it is difficult to detect and identify the vibration form of torsional vibration. The harm of torsional vibration to the shaft is not obvious in the early stages of torsional vibration, but the accumulation of torsional stress changes caused by torsional vibration often leads to accelerated wear of components, damage to couplings, shaft cracks and fractures, causing serious sudden accidents. Especially when the frequency of torsional vibration caused by external excitation is close to the natural frequency of the unit's shaft system, torsional resonance will occur, seriously affecting the safe operation of the unit. Based on the above accidents involving rotating shaft units, it is clear that torsional vibration can seriously affect the normal operation of industrial production and even cause huge property losses. Therefore, measuring and analyzing the torsional vibration of rotating machinery and realizing online real-time vibration monitoring of rotating machinery is of great significance for its stable operation and condition-based maintenance.

[0003] Methods for measuring torsional vibration signals can be broadly categorized into two types: torsional vibration measurement techniques based on the magnetoelectric pulse timing method and torsional vibration measurement techniques based on the photoelectric pulse timing method. The magnetoelectric pulse timing method obtains torsional vibration information of the shaft system by detecting changes in induced magnetic flux generated when a magnetoelectric sensor sweeps across the indexing structure on the rotating shaft. However, this method relies on the indexing structure mounted on the rotating shaft; under certain operating conditions, the shaft system is unsuitable for installing an indexing structure, and the installation of the indexing structure can also alter the structural characteristics of the shaft system. The photoelectric pulse timing method, on the other hand, obtains torsional vibration information of the shaft system by detecting high and low level signals generated when a laser pulse emitted by a fiber optic sensor sweeps across the zebra stripes pasted on the rotating shaft. It has the advantages of simple and convenient zebra stripe application and simple data acquisition equipment.

[0004] The torsional vibration signal measurement method based on fiber optic sensors can effectively collect torsional vibration signals, but the current technology is not yet mature in how to achieve real-time online monitoring of torsional vibration of rotating shaft systems under normal operating conditions of rotating equipment. Summary of the Invention

[0005] To address the problems of existing technologies, this invention provides a method for monitoring torsional vibration of rotating shaft systems based on fiber optic sensors. It estimates the real-time frequency of the torsional vibration signal using an ASA-LSF-based torsional vibration frequency estimation method and combines this with the ANSYS finite element method to obtain the natural torsional vibration frequency of the rotating shaft system for real-time monitoring. This invention eliminates reliance on prior knowledge such as signal processing techniques and diagnostic experience, offering simplicity and intuitiveness. Furthermore, it enables the diagnosis of torsional vibration in rotating machinery without disrupting normal operation, eliminating the need to shut down the machinery for testing. This provides a solution for avoiding torsional resonance in practical engineering applications.

[0006] The technical solution of this invention is as follows:

[0007] A method for monitoring torsional vibration of rotating shaft systems based on fiber optic sensors includes the following steps:

[0008] Step 1: Acquisition of torsional vibration signal based on fiber optic sensor;

[0009] The number of zebra stripes is determined according to the diameter of the rotating shaft system. The zebra stripes are then evenly pasted along the circumference of the rotating shaft system. The fiber optic sensor is installed perpendicular to the axis of the shaft system. Laser is emitted and received via a laser transmitter and receiver and the fiber optic sensor. Due to the different reflectivity of the zebra stripes to the laser, high and low level signals are generated after conversion by the photoelectric module. A high-frequency counter counts the number of pulses between two adjacent high levels, and after conversion, the time pulse interval between the two high levels is obtained. This data is then acquired by the host computer.

[0010] Step 2: Obtain the torsional natural frequency and real-time torsional frequency of the rotating shaft system.

[0011] The torsional natural frequencies of a rotating shaft system are obtained using modal analysis in ANSYS. Specifically, the shaft system is modeled according to its actual dimensions, meshed using the hexahedral method, and then imported into ANSYS Workbench for modal analysis. During modal analysis, the analysis order is set to the first eight orders to obtain the torsional natural frequencies and mode shapes of the corresponding orders.

[0012] The real-time frequency of torsional vibration is obtained using a non-stationary (ASA-LSF) torsional vibration signal frequency estimation method that combines angular domain synchronous averaging with least squares fitting. Specifically:

[0013] (a) Assume that the measured shaft system rotates one revolution and obtains m pulse intervals {T1, T2, ..., T...} m The sampling period is set to k, therefore the data length is m×k. The data is grouped according to the number of periods, and the following matrix C is constructed.

[0014]

[0015] The average value of matrix C by each column is shown below:

[0016]

[0017] To avoid torsional vibration calculation errors caused by rotor speed fluctuations, let the rotor period T be:

[0018]

[0019] The torsional displacement matrix S of all signal trigger points is obtained, in degrees:

[0020]

[0021] Therefore, the elements of matrix S are reordered to S':

[0022]

[0023] Formula (5) is the time-domain amplitude sequence.

[0024] (b) When calculating the instantaneous relative torsional vibration of the two mass disks, it is necessary to eliminate the black and white width error at the junction of the black and white stripes. Therefore, the least squares method is used to process the data. By observing the physical stripes, it can be determined that the signal at the stripe junction should be the maximum or minimum time interval value of any one period. Assume the average time interval value at the stripe junction is...

[0025] Selecting a portion of data from a certain period Where x i This is a time-domain numerical sequence, with units in radians, y i The dataset obtained from formula (5) is in degrees. Considering the actual computational cost, the polynomial is set to sixth order, and the relationship between the two is expressed as:

[0026]

[0027] Let the least squares loss function be:

[0028]

[0029] For a j Find the partial derivative and set it to 0, that is:

[0030]

[0031] Formula (8) yields the minimum parameter values ​​a1, a2, ..., a6 of the function.

[0032] When i = 1, the time interval value The corresponding predicted torsional displacement value is T1 1' =y1.

[0033] And so on, calculation Finally, it was concluded that... Recalculate the torsional displacement of the corresponding data points as follows: Replace matrix S By corresponding elements, a new torsional displacement matrix S' is obtained:

[0034]

[0035] By concatenating the first and last elements of each row in matrix S', we obtain:

[0036]

[0037] The sequence of elements within S” is used as the ordinate amplitude of the torsional vibration time-domain spectrum. As the step size, a time-domain spectrum is plotted, and the real-time spectrum of torsional vibration is calculated through FFT transformation. The horizontal axis of the spectrum is frequency in Hertz, and the vertical axis is torsional displacement in degrees. The instantaneous relative torsional displacement time-domain spectrum is obtained by subtracting the time-domain spectra of the two mass disks.

[0038] Step 3: Compare the real-time torsional vibration frequency with the natural frequency of the shaft system's torsional vibration, and compare the relative torsional vibration;

[0039] The difference between the real-time torsional vibration frequency obtained in step 2 and the natural torsional vibration frequency is compared. If the difference between the real-time torsional vibration frequency and the natural torsional vibration frequency of the shaft system is within the set threshold, and the relative fluctuation of the torsional displacement spectrum is large, it is considered that torsional vibration resonance is likely to occur, and a resonance warning is issued. If the difference between the real-time torsional vibration frequency and the natural torsional vibration frequency of the shaft system is outside the set threshold, it is considered that torsional vibration resonance is unlikely to occur, and no resonance warning is issued.

[0040] The beneficial effects of this invention are as follows: It utilizes fiber optic sensors to accurately acquire torsional vibration signals, and combines this with the ASA-LSF torsional vibration frequency estimation method to obtain more accurate real-time torsional vibration spectrum information. ANSYS finite element simulation software is used to obtain the torsional vibration natural frequencies of different rotating machinery shaft systems. Torsional vibration monitoring is performed by comparing the real-time torsional vibration spectrum with the range of the rotating shaft system's torsional vibration natural frequencies. This method effectively avoids cumbersome signal processing procedures and eliminates the reliance on prior knowledge such as signal processing techniques and fault diagnosis experience in signal waveform analysis methods. This invention can diagnose torsional vibration information without affecting the normal operation of rotating machinery; it is simple, intuitive, and convenient to use, and can achieve real-time monitoring of torsional vibration resonance. Attached Figure Description

[0041] Figure 1 is a flowchart of the rotating shaft torsional vibration monitoring method based on fiber optic sensor provided by the present invention;

[0042] Figure 2(a) and Figure 2(b) are a comparison of the spectrum obtained by processing real-time torsional vibration data using the torsional vibration frequency estimation method based on ASA-LSF proposed in this invention and the result of ordinary spectrum analysis calculation using the same set of data. Figure 2(a) is the FFT spectrum without the estimation algorithm, and Figure 2(b) is the FFT spectrum after the estimation algorithm is applied.

[0043] Figure 3 shows the modeling and mesh generation results of the torsional vibration test bench based on ANSYS in this invention.

[0044] Figure 4 shows the results of the first 8 modal analyses of the torsional vibration test bench based on ANSYS according to the present invention. Detailed Implementation

[0045] The specific embodiments of the present invention are described in detail below with reference to the technical solutions and accompanying drawings.

[0046] This embodiment presents a method for monitoring torsional vibration of rotating shafts based on fiber optic sensors, and the specific process is shown in Figure 1.

[0047] The data in this embodiment comes from a torsional vibration testing bench, which mainly consists of a torsional vibration generation system and a data acquisition system. The torsional vibration generation system primarily comprises a drive motor, a torsional vibration motor, a steel shaft, mass disks, a drive motor power supply, a drive motor speed controller, and an FK-308S frequency converter. The steel shaft of the rotating shaft system is made of solid steel, with a diameter of 10mm and a length of 720mm. Two mass disks, made of 45# steel and with a diameter of 78mm, are mounted on the rotating shaft system of the testing bench. Zebra stripe stickers with 35 stripes are attached to the mass disks. The data acquisition system mainly consists of fiber optic sensors, a GWYS-2 type laser transmitter and receiver, SH68F-68F cables, a BNC2121 junction box, a main unit chassis that can embed a PCI-6602 counting and acquisition card, and torsional vibration monitoring software.

[0048] The specific steps for monitoring torsional vibration of rotating shaft systems are as follows:

[0049] Step 1: Acquisition of torsional vibration signal based on fiber optic sensor;

[0050] The number of zebra stripes is determined according to the diameter of the rotating shaft system. The zebra stripes are then evenly pasted along the circumference of the rotating shaft system. The fiber optic sensor is installed perpendicular to the axis of the shaft system. Laser is emitted and received via a laser transmitter and receiver and the fiber optic sensor. Due to the different reflectivity of the zebra stripes to the laser, high and low level signals are generated after conversion by the photoelectric module. A high-frequency counter counts the number of pulses between two adjacent high levels, and after conversion, the time pulse interval between the two high levels is obtained. Data is then acquired via a data acquisition card.

[0051] Step 2: Use ANSYS modal analysis to obtain the torsional natural frequency of the rotating shaft system and use the ASA-LSF torsional vibration signal frequency estimation method to obtain the real-time torsional vibration frequency.

[0052] The model was created based on the actual dimensions of the shaft system, and the mesh was generated using the hexahedral method, as shown in Figure 3. The model was then imported into ANSYS Workbench for modal analysis. The analysis order was set to the first eight orders, and the natural frequencies and mode shapes for each order were obtained, as shown in Figure 4.

[0053] K periods of data were selected, a matrix was constructed, and then angular domain synchronous averaging was performed to obtain the averaged torsional vibration data. The averaged torsional vibration data was then subjected to least-squares fitting to obtain the values ​​of the zebra stripe start and end points. These fitted values ​​replaced the angular domain synchronously averaged data. Finally, a fast FFT transform was performed to obtain the real-time spectrum of the torsional vibration signal. The horizontal axis of the spectrum represents frequency in Hertz, and the vertical axis represents torsional angular displacement in degrees. Figures 2(a) and 2(b) show a comparison of the spectrum analysis with and without the frequency estimation algorithm when the torsional vibration frequency is 270Hz.

[0054] Step 3: Compare the real-time torsional vibration frequency with the natural torsional vibration frequency of the shaft system;

[0055] The system reads the natural torsional vibration frequency of the shaft system; calculates the torsional vibration signal frequency in real time; and compares the torsional vibration frequency at the point of maximum amplitude in the index spectrum with the natural torsional vibration frequency. If the difference between the real-time torsional vibration frequency and the natural torsional vibration frequency of the shaft system is within 2 Hz, and the relative torsional displacement spectrum fluctuates significantly, torsional resonance is considered likely to occur, and the system issues a resonance warning. If the difference between the real-time torsional vibration frequency and the natural torsional vibration frequency of the shaft system is more than 2 Hz, torsional resonance is considered unlikely to occur, and the system does not issue a resonance warning.

Claims

1. A method for monitoring torsional vibration of rotating shaft systems based on fiber optic sensors, characterized in that, The process includes the following steps: Step 1: Acquiring torsional vibration signals based on fiber optic sensors; determining the number of zebra stripes according to the diameter of the rotating shaft system, uniformly pasting zebra stripes along the circumference of the rotating shaft system, and installing the fiber optic sensor perpendicular to the axis of the shaft system; emitting and receiving laser light through a laser transmitter and receiver and the fiber optic sensor. Due to the different reflectivity of the zebra stripes to the laser light, high and low level signals are generated after conversion by the photoelectric module. A high-frequency counter is used to count the number of pulses between two adjacent high levels. After conversion, the time pulse interval between the two high levels is obtained, and the data is acquired by the host computer; Step 2: Obtaining the torsional vibration natural frequency and real-time frequency of the rotating shaft system. The torsional vibration natural frequency of the rotating shaft system is obtained using modal analysis in ANSYS. Specifically, the shaft system is modeled according to its actual dimensions, meshed using the hexahedral method, and imported into ANSYS. Modal analysis is performed in Workbench; during modal analysis, the analysis order is set to the first eight orders to obtain the torsional vibration natural frequencies and mode shapes of the corresponding orders; the real-time torsional vibration frequency is obtained using a non-stationary torsional vibration signal frequency estimation method that combines angular domain synchronous averaging with least squares fitting, specifically: (a) Assume that the measured shaft system rotates one revolution to obtain m pulse intervals {T1,T2,…,T…} m }, set the sampling period to k, so the data length is m×k; group the data according to the number of periods and construct the following matrix C; The average value of matrix C by each column is shown below: To avoid torsional vibration calculation errors caused by rotor speed fluctuations, let the rotor period T be: The torsional displacement matrix S of all signal trigger points is obtained, in degrees: Therefore, the elements of matrix S are reordered to S': Formula (5) is the time-domain amplitude sequence; (b) When calculating the instantaneous relative torsional vibration of the two mass disks, it is necessary to eliminate the black and white width error at the junction of the black and white stripes. Therefore, the least squares method is used to process the data. By observing the physical fringes, the signal at the fringe junction is determined to be either the maximum or minimum time interval value of any given period; assuming the average time interval value at the fringe junction is... Selecting a portion of data from a certain period Where x i This is a time-domain numerical sequence, with units in radians, y i This is a partial dataset obtained from formula (5), in degrees; Considering the actual computational complexity, the polynomial is defined as sixth order, and the relationship between the two is expressed as: Let the least squares loss function be: For a j Find the partial derivative and set it to 0, that is: Formula (8) yields the minimized parameter values ​​a1, a2, ..., a6 of the function; when i = 1, the time interval value... The corresponding predicted value of torsional displacement T1 1' For T1 1' =y1; and so on, calculate Finally, we obtain {T1} 1' T2 1' ,...,T k 1' }, recalculate the torsional displacement of the corresponding data points as follows Replace matrix S Corresponding elements yield a new torsional displacement matrix S': By concatenating the first and last elements of each row in matrix S', we obtain: The sequence of elements within S” is used as the ordinate amplitude of the torsional vibration time-domain spectrum. As the step size, a time-domain spectrum is plotted, and a real-time spectrum of torsional vibration is calculated using a fast FFT transform. The horizontal axis of the spectrum represents frequency in Hertz, and the vertical axis represents torsional displacement in degrees. The instantaneous relative torsional displacement time-domain spectrum is obtained by subtracting the time-domain spectra of the two mass disks. Step 3: Compare the real-time frequency of torsional vibration with the natural frequency of shaft torsional vibration, and compare the relative torsional vibration. The difference between the real-time frequency of torsional vibration obtained in Step 2 and the natural frequency of torsional vibration is compared. If the difference between the real-time frequency of torsional vibration and the natural frequency of shaft torsional vibration is within the set threshold, and the relative torsional displacement spectrum fluctuates significantly, it is considered that torsional resonance is likely to occur, and a resonance warning is issued. If the difference between the real-time frequency of torsional vibration and the natural frequency of shaft torsional vibration is outside the set threshold, it is considered that torsional resonance is unlikely to occur, and no resonance warning is issued.

Images