Engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics

Abstract

The present invention discloses an engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics. The method processes and analyzes engine gas path fault signals, calculates the wavelet packet entropy of the transmitter gas path fault signals in each frequency band, performs high-order statistics detection based on the wavelet packet entropy of each frequency band to obtain fault judgment characteristic values, and compares the fault judgment characteristic values ​​with standard thresholds to determine whether a fault occurs in the engine gas path. The present invention utilizes wavelet packet entropy to extract features of the engine gas path fault signals, can simultaneously decompose the low-frequency and high-frequency signals, and has higher time-frequency resolution. Then, high-order statistics are used to diagnose engine gas path faults on the extracted characteristic values, and the method has a high diagnostic accuracy for non-Gaussian and non-stationary fault signals, which has positive significance for subsequent health management such as life prediction.

Description

Technical Field

[0001] The invention belongs to the technical field of engine gas path detection and relates to an engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics. Background Art

[0002] Aircraft engine gas flow components, due to their complex structures and long-term operation under harsh operating conditions such as high temperature, high pressure, variable loads, and high stress, are highly susceptible to performance degradation and failures such as blade burns and fatigue damage, which can lead to failure of the entire equipment and cause significant losses. Gas flow components are widely recognized as a major source of failure in aircraft engines. The reliability of gas flow components not only affects engine efficiency and operating costs, but also equipment operational safety and personnel safety. According to statistics, in the past decade of flight accidents in my country, engine-related failures accounted for over 60% of mechanical and maintenance failures, of which gas flow component failures accounted for approximately 90%. These failures not only result in expensive repairs but also often lead to extremely serious accidents. Therefore, online monitoring of the condition of engine gas flow components is essential. Summary of the Invention

[0003] The purpose of the present invention is to provide an engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics, based on the wavelet packet entropy of the engine gas path fault, high-order statistics detection is performed based on the wavelet packet entropy of each frequency band to obtain a fault judgment characteristic value, and the fault judgment characteristic value is compared with the standard threshold to judge whether the engine gas path has a fault, which can be used to

[0004] The present invention is achieved through the following technical solutions:

[0005] An engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics is proposed. The engine gas path fault signal is processed and analyzed, and the wavelet packet entropy of the transmitter gas path fault signal in each frequency band is calculated. Based on the wavelet packet entropy of each frequency band, high-order statistics detection is performed to obtain the fault judgment characteristic value. The fault judgment characteristic value is then compared with the standard threshold to determine whether the engine gas path has a fault.

[0006] In order to better implement the present invention, further, the specific steps of performing high-order statistics detection based on the wavelet packet entropy of each frequency band to obtain the fault judgment characteristic value are as follows:

[0007] Step B1: Divide the wavelet packet entropy to be detected into K units, each unit contains M sample data, and then calculate the DFT coefficient of the wavelet packet entropy;

[0008] Step B2, calculating the bispectral intermediate variable of the triple correlation of the DFT coefficients;

[0009] Step B3, calculating the bispectrum according to the bispectrum intermediate variables;

[0010] Step B4: Calculate the fault judgment characteristic value based on the bispectrum;

[0011] Step B5: Determine the standard threshold value Th, and compare the fault judgment characteristic value T with the standard threshold value Th. If T≤Th, it is determined that the engine gas path is normal; if T>Th, it is determined that the engine gas path is faulty.

[0012] In order to better implement the present invention, further, the formula for calculating the DFT coefficient in step B1 is as follows:

[0013]

[0014] Of which: WPE i (λ) represents the wavelet packet entropy of the λth sample data in the i-th unit, j is the imaginary unit.

[0015] In order to better implement the present invention, further, the formula for calculating the bispectral intermediate variable in step B2 is as follows:

[0016]

[0017] in:

[0018] b i (λ1, λ2) is the intermediate variable of the bispectral unit of the i-th segment, 0≤λ2≤λ1, f s is the sampling frequency;

[0019] N0 is the number of sampling points; Δ0 is the frequency sampling interval; k1 and k2 are variables; i = 1, 2, ... K; λ1 and λ2 are variables.

[0020] In order to better implement the present invention, further, the formula for calculating the bispectrum in step B3 is as follows:

[0021]

[0022] Among them: B i (ω1, ω2) is the bispectrum of the i-th unit; b i (ω1, ω2) is the intermediate variable of the bispectral unit of the i-th segment; ω1, ω2 are angular frequencies, and In order to better implement the present invention, further, the formula for calculating the fault judgment characteristic value in step B4 is as follows:

[0023]

[0024] Where: T is the fault judgment characteristic value.

[0025] In order to better implement the present invention, further, the steps for calculating the wavelet packet entropy of each frequency band are as follows:

[0026] Step A1: Decompose the engine gas path fault signal using J-layer wavelet packets to obtain wavelet packet signals in 2J frequency bands;

[0027] Step A2, reconstructing the wavelet packet coefficients of the wavelet packet signals of 2J frequency bands to obtain 2J wavelet packet reconstructed signals;

[0028] Step A3, calculating the power spectrum energy of the wavelet packet reconstructed signal of the 2J frequency bands, and calculating the energy fraction of the wavelet packet reconstructed signal of the 2J frequency bands according to the power spectrum energy;

[0029] Step A4: Calculate the wavelet packet entropy of 2J frequency bands according to the energy fraction.

[0030] In order to better implement the present invention, further, the formula for reconstructing the wavelet packet coefficients in step A2 is as follows:

[0031]

[0032] in:

[0033] is the wavelet packet reconstruction signal of the i-th signal of the k-th node in the J-th layer; f(t) is the input wavelet packet signal; is the wavelet packet decomposition result of the i-th signal of the k-th node in the J-th layer; t is the time variable; i = 1, 2, ...N.

[0034] In order to better implement the present invention, further, the calculation formula of the power spectrum energy in step A3 is as follows:

[0035]

[0036] Where: E(k) is the power energy spectrum of the wavelet packet in the kth frequency band;

[0037] The calculation formula of the energy fraction in step A3 is as follows:

[0038]

[0039] Where: P k is the energy fraction of the wavelet packet in the kth frequency band.

[0040] In order to better implement the present invention, further, the calculation formula of the wavelet packet entropy in step A4 is as follows:

[0041]

[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0043] The present invention uses wavelet packet entropy to extract features from engine gas path fault signals, can simultaneously decompose the low-frequency and high-frequency signals, and has higher time-frequency resolution; then uses high-order statistics to diagnose engine gas path faults on the extracted eigenvalues, and has a high diagnostic accuracy for non-Gaussian and non-stationary fault signals, which has positive significance for subsequent health management such as life prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Schematic diagram of the process steps of the present invention. DETAILED DESCRIPTION

[0045] Example 1:

[0046] This embodiment's engine gas path fault diagnosis method, based on wavelet packet entropy and high-order statistics, processes and analyzes engine gas path fault signals. These signals are acquired by sensors using an engine gas path electrostatic detection device that applies static electricity to the engine gas path. The method calculates the wavelet packet entropy of the transmitter gas path fault signal in each frequency band. High-order statistics are then detected based on the wavelet packet entropy in each frequency band to obtain fault diagnosis eigenvalues. These eigenvalues ​​are then compared with standard thresholds to determine whether an engine gas path fault has occurred.

[0047] like Figure 1 As shown in FIG, the specific steps of performing high-order statistics detection based on the wavelet packet entropy of each frequency band to obtain the fault judgment characteristic value are as follows:

[0048] The calculation steps of wavelet packet entropy of each frequency band are:

[0049] Step A1: Decompose the engine gas path fault signal using J-layer wavelet packets to obtain wavelet packet signals in 2J frequency bands;

[0050] Step A2, reconstructing the wavelet packet coefficients of the wavelet packet signals of 2J frequency bands to obtain 2J wavelet packet reconstructed signals;

[0051] Step A3, calculating the power spectrum energy of the wavelet packet reconstructed signal of the 2J frequency bands, and calculating the energy fraction of the wavelet packet reconstructed signal of the 2J frequency bands according to the power spectrum energy;

[0052] Step A4: Calculate the wavelet packet entropy of 2J frequency bands according to the energy fraction.

[0053] Step B1: Divide the wavelet packet entropy to be detected into K units, each unit contains M sample data, and then calculate the DFT coefficient of the wavelet packet entropy;

[0054] Step B2, calculating the bispectral intermediate variable of the triple correlation of the DFT coefficients;

[0055] Step B3, calculating the bispectrum according to the bispectrum intermediate variables;

[0056] Step B4: Calculate the fault judgment characteristic value based on the bispectrum;

[0057] Step B5: Determine the standard threshold value Th, and compare the fault judgment characteristic value T with the standard threshold value Th. If T≤Th, it is determined that the engine gas path is normal; if T>Th, it is determined that the engine gas path is faulty.

[0058] The formula for calculating the DFT coefficients in step B1 is as follows:

[0059]

[0060] in: represents the wavelet packet entropy of the λth sample data in the i-th segment unit, j is the imaginary unit.

[0061] The formula for calculating the bispectral intermediate variable in step B2 is as follows:

[0062]

[0063] in:

[0064] b i (λ1, λ2) is the intermediate variable of the bispectral unit of the i-th segment, and 0≤λ2≤λ1, f s is the sampling frequency;

[0065] N0 is the number of sampling points; Δ0 is the frequency sampling interval; k1 and k2 are variables; i = 1, 2, ... K; λ1 and λ2 are variables.

[0066] The formula for calculating the bispectrum in step B3 is as follows:

[0067]

[0068] Among them: B i (ω1, ω2) is the bispectrum of the i-th unit; b i (ω1, ω2) is the intermediate variable of the bispectral unit of the i-th segment; ω1, ω2 are angular frequencies, and

[0069] The formula for calculating the fault judgment characteristic value in step B4 is as follows:

[0070]

[0071] Where: T is the fault judgment characteristic value.

[0072] The formula for reconstructing the wavelet packet coefficients in step A2 is as follows:

[0073]

[0074] in: is the wavelet packet reconstruction signal of the i-th signal of the k-th node in the J-th layer; f(t) is the input wavelet packet signal; is the wavelet packet decomposition result of the i-th signal of the k-th node in the J-th layer; t is the time variable; i = 1, 2, ...N.

[0075] The calculation formula of the power spectrum energy in step A3 is as follows:

[0076]

[0077] Where: E(k) is the power energy spectrum of the wavelet packet in the kth frequency band;

[0078] The calculation formula of the energy fraction in step A3 is as follows:

[0079]

[0080] Where: Pk is the energy fraction of the wavelet packet in the kth frequency band.

[0081] The calculation formula of the wavelet packet entropy in step A4 is as follows:

[0082]

[0083] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the scope of protection of the present invention.

Claims

1. An engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics processes and analyzes engine gas path fault signals, which is characterized by: Calculate the wavelet packet entropy of the transmitter gas path fault signal in each frequency band, perform high-order statistical detection based on the wavelet packet entropy of each frequency band to obtain a fault judgment characteristic value, and compare the fault judgment characteristic value with a standard threshold to determine whether a fault has occurred in the engine gas path; wherein the specific steps of performing high-order statistical detection based on the wavelet packet entropy of each frequency band to obtain the fault judgment characteristic value are as follows: Step B1: Divide the wavelet packet entropy to be detected into K units, each unit contains M sample data, and then calculate the DFT coefficient of the wavelet packet entropy; Step B2: Calculate the bispectral intermediate variable of the triple correlation of the DFT coefficients; the formula for calculating the bispectral intermediate variable in step B2 is as follows: Where: b i (λ1, λ2) is the intermediate variable of the bispectral unit of the i-th segment, and 0≤λ2≤λ1, f s is the sampling frequency; N0 is the number of sampling points; Δ0 is the frequency sampling interval; k1 and k2 are variables; i = 1, 2, ... K; λ1 and λ2 are variables; Step B3, calculating the bispectrum according to the bispectrum intermediate variable; the formula for calculating the bispectrum in step B3 is as follows: Among them: B i (ω1, ω2) is the bispectrum of the i-th unit; b i (ω1, ω2) is the intermediate variable of the bispectral unit of the i-th segment; ω1, ω2 are angular frequencies, and Step B4: Calculate the fault judgment characteristic value based on the bispectrum; Step B5: Determine the standard threshold value Th, and compare the fault judgment characteristic value T with the standard threshold value Th. If T≤Th, it is determined that the engine gas path is normal; if T>Th, it is determined that the engine gas path is faulty.

2. The engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics according to claim 1 is characterized in that: The formula for calculating the DFT coefficients in step B1 is as follows: Of which: WPE i (λ) represents the wavelet packet entropy of the λth sample data in the i-th unit, n=1, 2,…, M-1; j is an imaginary unit.

3. The engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics according to claim 1, characterized in that: The formula for calculating the fault judgment characteristic value in step B4 is as follows: Where: T is the fault judgment characteristic value.

4. The engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics according to claim 1, characterized in that: The calculation steps of wavelet packet entropy of each frequency band are: Step A1: Decompose the engine gas path fault signal by J-layer wavelet packet to obtain 2 J Wavelet packet signal of frequency band; Step A2, pair 2 J The wavelet packet signal of the frequency band is reconstructed by wavelet packet coefficients to obtain 2 J Wavelet packets reconstruct the signal; Step A3, Calculation 2 J The power spectrum energy of the wavelet packet reconstructed signal in each frequency band is calculated based on the power spectrum energy. J The energy fraction of the wavelet packet reconstructed signal in each frequency band; Step A4: Calculate 2 based on energy fraction J The wavelet packet entropy of each frequency band.

5. The engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics according to claim 4 is characterized in that: The formula for reconstructing the wavelet packet coefficients in step A2 is as follows: in: is the wavelet packet reconstruction signal of the i-th signal of the k-th node in the J-th layer; f(t) is the input wavelet packet signal; is the wavelet packet decomposition result of the i-th signal of the k-th node in the J-th layer; t is the time variable; i = 1, 2, ...N.

6. The engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics according to claim 4, characterized in that: The calculation formula of the power spectrum energy in step A3 is as follows: Where: E(k) is the power energy spectrum of the wavelet packet in the kth frequency band; The calculation formula of the energy fraction in step A3 is as follows: Where: P k is the energy fraction of the wavelet packet in the kth frequency band.

7. The engine gas path fault diagnosis method based on wavelet packet entropy and high-order statistics according to claim 4, characterized in that: The calculation formula of the wavelet packet entropy in step A4 is as follows:

Images