Turbine blade noise analysis and diagnosis method based on welch algorithm

Abstract

The invention provides a turbine blade noise analysis and diagnosis method based on a welch algorithm, and relates to the field of aerodynamic noise analysis and control. The method comprises the following steps: a) acquiring a turbine pulsating pressure signal and a noise time signal by using a piezoresistive high-frequency dynamic pressure sensor and a microphone array; b) performing high-pass filtering de-trending, segmented windowing and fast Fourier transform on the turbine pulsating pressure signal and the noise time signal, and calculating an average power spectrum according to a Welch algorithm; and c) blade faults are diagnosed according to the average power spectrum and the health reference power spectrum density under the same working condition. According to the method, the noise change can be controlled in real time, so that whether the blade breaks down or not is judged.

Description

Technical Field

[0001] This invention relates to the field of aerodynamic noise analysis and control, and in particular to a turbine blade noise analysis and diagnosis method based on the Welch algorithm. Background Technology

[0002] With the improvement of economic level and the progress of science and technology, aerodynamic noise is becoming increasingly closely related to people's lives. The proportion of research on the generation and propagation mechanism of various aerodynamic noises during aircraft flight and the development of effective noise reduction measures is increasing year by year in aviation industry research and development. Many aviation research institutions at home and abroad have listed the control of aerodynamic noise as one of the main directions of aviation technology development.

[0003] Turbine blade failures can lead to abnormal mechanical vibrations, increased airflow disturbances, or compromised structural integrity, significantly altering the noise characteristics of the turbine system. Noise analysis technology can determine in real time whether a blade failure has occurred and accurately pinpoint the location of the fault. Furthermore, combining aerodynamic optimization and intelligent control can effectively suppress noise and prevent cascading failures.

[0004] Traditional noise analysis methods (such as A-weighted sound pressure level measurement) struggle to distinguish the physical sources of broadband and narrowband noise. Spectral analysis often neglects the correlation between fluid parameters (velocity, pressure) and noise, resulting in a lack of targeted optimization measures. The vortex-acoustic-structural coupling effect in turbomachinery is complex, and existing methods cannot effectively separate the contributions of multiple physics fields. Analyzing the dynamic correlation between frequency domain energy distribution characteristics and hydrodynamic parameters can achieve high-precision noise source localization and noise reduction optimization. PSD calculations using the Welch algorithm on acquired pressure and noise signals demonstrate stronger noise resistance compared to conventional periodogram methods, producing smoother PSD curves that are more suitable for steady-state turbomachinery. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a turbine blade noise analysis and diagnosis method based on the Welch algorithm. This method can analyze the dynamic correlation between frequency domain energy distribution characteristics and hydrodynamic parameters, enabling real-time control of whether a blade fault has occurred and precise location of the fault. Furthermore, by combining aerodynamic optimization and intelligent control, it can effectively suppress noise and prevent cascading failures.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A turbine blade noise analysis and diagnosis method based on the Welch algorithm includes:

[0008] a) Use a piezoresistive high-frequency dynamic pressure sensor and microphone array to acquire turbine pulsating pressure signal and noise time signal;

[0009] b) Perform high-pass filtering to de-trend the turbine pulsating pressure signal and the noise time signal, segment windowing and fast Fourier transform, and calculate the average power spectrum according to the Welch algorithm;

[0010] c) Diagnose blade faults based on the average power spectrum and the power spectral density of the healthy reference under the same operating conditions.

[0011] Preferably, a piezoresistive high-frequency dynamic pressure sensor and a microphone array are used to acquire turbine pulsating pressure signals and noise time signals, including:

[0012] The piezoresistive high-frequency dynamic pressure sensor is installed 5-10 cm in front of the turbine inlet guide vane and 5-10 cm behind the trailing edge of the last stage moving blade, with the diaphragm flush with the flow channel wall, to simultaneously measure the turbine pulsating pressure signal at the inlet, outlet and in the channel.

[0013] A microphone array is uniformly arranged on the flow channel wall between the trailing edge of each stationary blade and the leading edge of the moving blade to collect the noise time signal at the corresponding spatial location.

[0014] Preferably, the turbine pulsating pressure signal and the noise time signal are subjected to high-pass filtering for detrending, piecewise windowing, and fast Fourier transform, and the average power spectrum is calculated according to the Welch algorithm, including:

[0015] The turbine pulsating pressure signal and the noise time signal acquired synchronously are respectively subjected to high-pass filtering to remove trend;

[0016] Each signal after high-pass filtering and descaling is segmented, and a Hanning window is applied to each segment.

[0017] Perform a Fast Fourier Transform on each windowed signal segment to obtain the corresponding frequency domain signal;

[0018] The average power spectrum is obtained by averaging the frequency domain signals of each segment using the Welch algorithm.

[0019] Preferably, during the pass-through filtering de-stressing process, the cutoff frequency of the filter is within the range of the turbine shaft frequency. to between.

[0020] Preferably, the filter is an FIR filter, and the filter order is set to 6th order.

[0021] Preferably, the length of the Hanning window is 1024 points, and the overlap rate is 50%.

[0022] Preferably, diagnosing blade faults based on the average power spectrum and the power spectral density of a healthy reference under the same operating conditions includes:

[0023] The average power spectrum is compared with the power spectrum density of the health benchmark under the same operating conditions. If the difference obtained by the comparison does not exceed the preset threshold, the process returns to step a) to continue monitoring. If the difference exceeds the threshold, the blade fault is determined based on the abnormal spectrum segment and a shutdown inspection command is output.

[0024] Preferably, the blade failure type includes at least cracks, wear, or surface damage.

[0025] The present invention discloses the following technical effects:

[0026] 1. This invention proposes a turbine blade noise analysis and diagnosis method based on the Welch algorithm. By dynamically correlating the frequency domain energy distribution characteristics with fluid dynamic parameters, and combining CFD simulation data with experimental dynamic pressure signals, a mapping relationship between noise energy spectrum and flow field characteristics is constructed to achieve real-time control of noise changes, thereby determining whether the blade has failed.

[0027] 2. This invention performs high-pass filtering to remove the trend term from the acquired time-domain signal. High-pass filtering can more accurately separate the trend term from the effective signal frequency band, thereby improving the accuracy of spectrum analysis.

[0028] 3. The PSD calculation of the collected pressure and noise signals using the Welch algorithm has stronger noise resistance than the conventional periodogram method, and the calculated PSD curve is smoother, making it more suitable for turbomachinery operating in steady state. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 A flowchart of the method provided in an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of the technical route provided for an embodiment of the present invention. Detailed Implementation

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0033] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] Figure 1 A flowchart of the method provided in the embodiments of the present invention, such as Figure 1 As shown, this invention provides a turbine blade noise analysis and diagnosis method based on the Welch algorithm, including:

[0035] a) Use a piezoresistive high-frequency dynamic pressure sensor and microphone array to acquire turbine pulsating pressure signal and noise time signal;

[0036] b) Perform high-pass filtering, segmented windowing, and fast Fourier transform on the turbine pulsating pressure signal and noise time signal, and calculate the average power spectrum according to the Welch algorithm;

[0037] c) Diagnose blade faults based on the average power spectrum and the power spectral density of the healthy baseline under the same operating conditions.

[0038] like Figure 2 As shown, the technical approach of this embodiment is as follows:

[0039] (1) The piezoresistive high-frequency dynamic pressure sensor is placed 5-10 cm in front of the turbine inlet guide vane to measure the inlet pressure signal, and 5-10 cm behind the trailing edge of the last stage moving blade to measure the outlet pressure signal. The diaphragm of the piezoresistive pressure sensor in the turbine channel is installed flush with the flow channel wall to measure the pulsating pressure signal in the turbine channel. The flush installation method can minimize the influence of the sensor on the pressure signal. The microphone array is evenly distributed on the flow channel wall between the stationary blades and the moving blades to collect the noise signal in the turbine. Placing the microphone array between the trailing edge of the stationary blade and the leading edge of the moving blade can better collect the noise signal here, which is more conducive to analyzing the influence of flow separation and vortex structure at the trailing edge of the stationary blade on turbine noise.

[0040] (2) Remove trend terms: Use high-pass filtering to remove low-frequency trend terms.

[0041] Specifically, the acquired signal is detrended to obtain signal x{n}, and a high-pass filter is used for detrending. With the turbine speed set to 3000 rpm, the cutoff frequency f is... c The frequency range should be 5-10Hz. Use an FIR filter with a filter order of 6. Select zero-phase filtering for phase processing. Zero-phase filtering can avoid the time delay of the surge signal.

[0042] (3) Segmented windowing: Segment the long signal to reduce spectrum leakage.

[0043] Specifically, the data after detrending is segmented and windowed. The signal x{n} (of length N) is divided into M segments, each of length L, where L is the window length. Adjacent segments overlap by O points, and the overlap rate is... Then the m-th segment of the signal can be represented as:

[0044] x m {n}=x[n+(m-1)(LO)],n=0,1,…,L-1(m=1,2,…,M)

[0045] Let the highest frequency of the original signal be f. max If the frequency is 50kHz, then the sampling rate f s.max For 128kHz, take L = 1024, OL = 50% (O = 512), M ≈ N / (LO).

[0046] The Hanning window function is used for each segment of signal x. m Windowing is applied to {n} to obtain signal x. m，win {n}, where the Hanning window function is:

[0047] w{n}=0.5-0.5cos(2πn / (L-1))

[0048] (4) For each windowed signal x m，win The frequency domain signal X is obtained by performing a Fourier transform on {n}. m {k}

[0049]

[0050] For each frequency domain signal X m {k} Calculate the single-segment power spectrum P m (f k ):

[0051] Specifically,

[0052] in, Frequency resolution

[0053] The average power spectrum is obtained by averaging the power spectra of all segments.

[0054] Specifically,

[0055] The heavy-duty gas turbine high-pressure stage operates at 3000 rpm. The first stage has 42 stationary blades and 84 moving blades. Therefore, the passing frequencies of the moving and stationary blades are:

[0056]

[0057] Extract the PSD curve within 6 times the pass frequency, i.e., 0-25200Hz.

[0058] (5) Perform spectrum analysis on the processed signal.

[0059] (6) Compare the analysis results with the health status, and perform the following operations based on the comparison results:

[0060] If the noise PSD is not significantly different from that in the healthy state, repeat the above acquisition and analysis steps.

[0061] If there is a significant difference in noise PSD, the fault location should be determined based on the spectral difference, and the machine should be stopped to inspect the blades.

[0062] Specifically, if a high-frequency resonance peak appears, it indicates that the blade has a loose, cracked or broken fault, which causes the blade to resonate. The broken fragments hit the shell or adjacent blades, generating instantaneous impact noise.

[0063] If the fundamental frequency is equal to the rotational speed frequency, it indicates that the blade has a fault such as wear or deformation. This fault leads to uneven distribution of centrifugal force, causing the blade to rotate eccentrically. The airflow periodically sweeps the blade gap, generating pulsating pressure, and the noise amplitude increases linearly with the rotational speed until saturation.

[0064] If broadband turbulent noise occurs, it indicates that the blades have surface damage, such as corrosion or erosion. Increased surface roughness or blockage of cooling holes leads to increased airflow separation and vortex shedding, resulting in increased turbulent noise energy.

[0065] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0066] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for analyzing and diagnosing turbine blade noise based on the Welch algorithm, characterized in that, include: a) Use a piezoresistive high-frequency dynamic pressure sensor and microphone array to acquire turbine pulsating pressure signal and noise time signal; b) Perform high-pass filtering to de-trend the turbine pulsating pressure signal and the noise time signal, segment windowing and fast Fourier transform, and calculate the average power spectrum according to the Welch algorithm; c) Diagnose blade faults based on the average power spectrum and the power spectral density of the healthy reference under the same operating conditions.

2. The turbine blade noise analysis and diagnosis method based on the Welch algorithm according to claim 1, characterized in that, Using a piezoresistive high-frequency dynamic pressure sensor and a microphone array, turbine pulsating pressure signals and noise time signals are acquired, including: The piezoresistive high-frequency dynamic pressure sensor is installed 5-10 cm in front of the turbine inlet guide vane and 5-10 cm behind the trailing edge of the last stage moving blade, with the diaphragm flush with the flow channel wall, to simultaneously measure the turbine pulsating pressure signal at the inlet, outlet and in the channel. A microphone array is uniformly arranged on the flow channel wall between the trailing edge of each stationary blade and the leading edge of the moving blade to collect the noise time signal at the corresponding spatial location.

3. The turbine blade noise analysis and diagnosis method based on the Welch algorithm according to claim 1, characterized in that, The turbine pulsating pressure signal and the noise time signal are subjected to high-pass filtering for detrending, piecewise windowing, and fast Fourier transform, and the average power spectrum is calculated according to the Welch algorithm, including: The turbine pulsating pressure signal and the noise time signal acquired synchronously are respectively subjected to high-pass filtering to remove trend; Each signal after high-pass filtering and descaling is segmented, and a Hanning window is applied to each segment. Perform a Fast Fourier Transform on each windowed signal segment to obtain the corresponding frequency domain signal; The average power spectrum is obtained by averaging the frequency domain signals of each segment using the Welch algorithm.

4. The turbine blade noise analysis and diagnosis method based on the Welch algorithm according to claim 1, characterized in that, During the pass-through filtering de-stressing process, the filter cutoff frequency ranges from the turbine shaft frequency. to between.

5. The turbine blade noise analysis and diagnosis method based on the Welch algorithm according to claim 4, characterized in that, The filter is an FIR filter, and its order is set to 6.

6. The turbine blade noise analysis and diagnosis method based on the Welch algorithm according to claim 3, characterized in that, The length of the Hanning window is 1024 points, with an overlap rate of 50%.

7. The turbine blade noise analysis and diagnosis method based on the Welch algorithm according to claim 1, characterized in that, Blade fault diagnosis is performed based on the average power spectrum and the power spectral density of a healthy reference under the same operating conditions, including: The average power spectrum is compared with the power spectrum density of the health benchmark under the same operating conditions. If the difference obtained by the comparison does not exceed the preset threshold, the process returns to step a) to continue monitoring. If the difference exceeds the threshold, the blade fault is determined based on the abnormal spectrum segment and a shutdown inspection command is output.

8. The turbine blade noise analysis and diagnosis method based on the Welch algorithm according to claim 1, characterized in that, The blade failure types include at least cracks, wear, or surface damage.

Images