A method for automatically and quickly processing dynamic stress test data

Abstract

The application provides a kind of dynamic stress test data automatic fast processing method, comprising the following steps: ① distinguish signal type: according to the change of original signal, judge the type of the test section where current data is located;Original signal includes speed data and vibration data;② adjust time-frequency resolution: according to the type of test section, set time resolution and frequency resolution, and time resolution Δt and frequency resolution Δf satisfy condition ③ transform calculation: the original signal is processed using short-time Fourier transform.The application can quickly and accurately analyze and process the measured vibration signal, obtain key vibration data, obtain dynamic frequency-speed diagram to determine the stress-speed characteristic of blade in the working speed range, determine whether there is dangerous resonance in the engine operating speed range, and complete the blade theoretical calculation strength and actual measurement strength checking.

Description

Technical Field

[0001] This invention relates to an automatic and rapid processing method for dynamic stress test data. Background Technology

[0002] Currently, commonly used dynamic data acquisition systems employ a pre-acquisition, post-processing approach for acquiring and analyzing dynamic stress test data of blades. Furthermore, the results obtained through spectral analysis cannot directly provide numerical values ​​such as order, rotational speed, frequency, and vibration amplitude. For non-stationary vibration signals generated during non-stationary processes such as acceleration and deceleration in turbine engines, the use of general Fourier transforms can produce a "spectral ambiguity" phenomenon, easily leading to significant errors in the analysis results.

[0003] The methods used by dynamic signal analyzers in the market to analyze and process the vibration signals of acquired rotating machinery mainly involve simultaneously acquiring rotational speed pulse signals and vibration signals, calculating the rotational speed curve from the rotational speed pulse signals (with time on the x-axis and rotational speed on the y-axis), and using the rotational speed curve as the basis for vibration signal analysis. Then, the rotational speed curve and vibration signal are manually analyzed synchronously to obtain the spectral data of the vibration signal. A color map and waterfall plot are created using the spectral data, and finally, the blade vibration frequency, order, and engine speed corresponding to the moments with larger blade vibration amplitudes are extracted from the spectrum, completing the processing of the experimental data.

[0004] Traditional methods for processing dynamic stress measurement test data of aero-engine blades mainly involve performing spectrum transformation on test data with the same frequency resolution throughout the entire start-up process. This results in the inability to clearly distinguish the main vibration frequency of the blade as the engine speed changes and the magnitude of the stress on the blade from the vibration signals measured under two completely different states: steady-state vibration and rapid vibration. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an automatic and rapid processing method for dynamic stress test data. This method can quickly and accurately analyze and process the measured vibration signals to obtain key vibration data, acquire a dynamic frequency-speed diagram (Campbell diagram) to determine the stress-speed characteristics of the blade within the operating speed range, determine whether there is any dangerous resonance in the blade within the engine's operating speed range, and complete the verification between the theoretical calculated strength and the actual measured strength of the blade.

[0006] The present invention is achieved through the following technical solutions.

[0007] This invention provides an automatic and rapid processing method for dynamic stress test data, comprising the following steps:

[0008] ① Distinguish signal types: Based on the changes in the original signal, determine the type of test section the current data belongs to; the original signal includes rotational speed data and vibration data;

[0009] ② Adjust time and frequency resolution: Set the time resolution and frequency resolution according to the type of test section, and ensure that the time resolution Δt and frequency resolution Δf meet the following conditions.

[0010] ③ Transformation calculation: The original signal is processed using the short-time Fourier transform.

[0011] The short-time Fourier transform formula is as follows:

[0012]

[0013] In the formula, g() is the window function, x(τ) is the original signal, * is the conjugate, τ is the center of the window function, and STFT(t,f) is a two-dimensional function with respect to time t and frequency f.

[0014] The short-time Fourier transform is obtained by discretizing a discrete time-frequency grid (mT, nF) with equal intervals using the following formula:

[0015]

[0016] In the formula, m, n, p, and q are all integers, representing different engine states automatically determined based on engine speed, and T... p F q These represent the different time resolutions and frequency resolutions used in each state, respectively, and satisfy T. p >0, F q >0, STFT(mT, nF) is a two-dimensional function of mT and nF.

[0017] In step ①, the following steps are specifically adopted:

[0018] S1.1. Denoising: Denoise the entire data segment of length L;

[0019] S1.2. Piecewise Fitting: A linear function is used to fit each segment of data to obtain the fitting equation for each segment;

[0020] S1.3. Calculate the mean: Calculate the average and root mean square of the slope of each data segment;

[0021] S1.4. Classification: Based on the mean and root mean square, the slope of each data segment is classified into types greater than the threshold and types less than the threshold. Types greater than the threshold include acceleration and deceleration states, while types less than the threshold include steady states.

[0022] S1.5. Specific classification: Differential equations are used to distinguish the data segments with values ​​greater than the threshold and to determine the type of test segment to which they belong.

[0023] In step S1.1, it is divided into n segments, n = L / L win Let each data segment be N1, N2, ..., N n .

[0024] In step S1.2, the fitting equation for each segment is Y. 0i =k Oi +b 0i (i = 1, 2, ..., n), if the coefficient of the linear term k 0i If the value is greater than or equal to zero, then the data segment is sorted in ascending order; if the coefficient of the linear term k is greater than or equal to zero... 0i If the value is less than zero, the data segment is sorted in descending order.

[0025] The rearranged data is then fitted again using a linear function, resulting in the fitting equation Y1i=k1i+b1i (i=1,2,...,n) for each data segment, where k1i represents the rate of change of rotational speed with time, i.e., the slope.

[0026] In step S1.3, the average slope of each data segment is:

[0027]

[0028] The root mean square of the slope of each data segment is:

[0029]

[0030] In step S1.4, the threshold is k. threshold =k mean +k std , where k mean k is the average slope of the data segments. std This represents the root mean square of the slope of each data segment.

[0031] In step S1.5, the difference equation is Y = [x(2)-x(1), x(3)-x(2), ..., x(n)-x(n-1)], the length of Y is n-1 and it is stored in array form. In array Y, -1 indicates that the process is the beginning of the deceleration state, and 1 indicates that the process is the beginning of the acceleration state.

[0032] The beneficial effects of this invention are: it can quickly and accurately analyze and process the measured vibration signals to obtain key vibration data, obtain a dynamic frequency-speed diagram (Campbell diagram) to determine the stress-speed characteristics of the blade in the operating speed range, determine whether there is dangerous resonance in the blade in the engine operating speed range, complete the verification of the theoretical calculated strength of the blade with the actual measured strength, and effectively reduce the time spent in data processing, ensuring that the experimental data processing is more accurate, convenient and efficient. Attached Figure Description

[0033] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0034] The technical solution of the present invention is further described below, but the scope of protection is not limited to what is described.

[0035] Example 1

[0036] like Figure 1 The method for automatic and rapid processing of dynamic stress test data, as shown, includes the following steps:

[0037] ① Distinguish signal types: Based on the changes in the original signal, determine the type of test section the current data belongs to; the original signal includes rotational speed data and vibration data;

[0038] ② Adjust time and frequency resolution: Set the time resolution and frequency resolution according to the type of test section, and ensure that the time resolution Δt and frequency resolution Δf meet the following conditions.

[0039] ③ Transformation calculation: The original signal is processed using the short-time Fourier transform.

[0040] Example 2

[0041] Based on Example 1, the short-time Fourier transform formula is:

[0042]

[0043] In the formula, g() is the window function, x(τ) is the original signal, * is the conjugate, τ is the center of the window function, and STFT(t,f) is a two-dimensional function with respect to time t and frequency f.

[0044] The short-time Fourier transform is obtained by discretizing the time-frequency grid (mT, nF) with equal intervals using the following formula:

[0045]

[0046] In the formula, m, n, p, and q are all integers, representing different engine states automatically determined based on engine speed, and T...p F q These represent the different time resolutions and frequency resolutions used in each state, respectively, and satisfy T. p >0, F q >0, STFT(mT, nF) is a two-dimensional function of mT and nF.

[0047] Example 3

[0048] Based on Example 1, step ① specifically adopts the following steps:

[0049] S1.1. Denoising: Denoise the entire data segment of length L;

[0050] S1.2. Piecewise Fitting: A linear function is used to fit each segment of data to obtain the fitting equation for each segment;

[0051] S1.3. Calculate the mean: Calculate the average and root mean square of the slope of each data segment;

[0052] S1.4. Classification: Based on the mean and root mean square, the slope of each data segment is classified into types greater than the threshold and types less than the threshold. Types greater than the threshold include acceleration and deceleration states, while types less than the threshold include steady states.

[0053] S1.5. Specific classification: Differential equations are used to distinguish the data segments with values ​​greater than the threshold and to determine the type of test segment to which they belong.

[0054] In step S1.1, it is divided into n segments, n = L / L win Let each data segment be N1, N2, ..., N n .

[0055] In step S1.2, the fitting equation for each segment is Y. 0i =k 0i +b 0i (i = 1, 2, ..., n), if the coefficient of the linear term k 0i If the value is greater than or equal to zero, then the data segment is sorted in ascending order; if the coefficient of the linear term k is greater than or equal to zero... 0i If the value is less than zero, the data segment is sorted in descending order.

[0056] The rearranged data is then fitted again using a linear function, resulting in the fitting equation Y1i=k1i+b1i (i=1,2,...,n) for each data segment, where k1i represents the rate of change of rotational speed with time, i.e., the slope.

[0057] In step S1.3, the average slope of each data segment is:

[0058]

[0059] The root mean square of the slope of each data segment is:

[0060]

[0061] In step S1.4, the threshold is k. threshold =k mean +k std , where k mean k is the average slope of the data segments. std This represents the root mean square of the slope of each data segment.

[0062] In step S1.5, the difference equation is Y = [x(2)-x(1), x(3)-x(2), ..., x(n)-x(n-1)]. The length of Y is n-1 and it is stored in array form. In array Y, -1 indicates that the process is the beginning of the deceleration state, and 1 indicates that the process is the beginning of the acceleration state.

[0063] Therefore, this invention solves the problem of rapid changes in vibration signals during the acceleration and deceleration of aero-engines, and provides analysis with good time resolution to obtain the frequency of blade vibration and the magnitude of dynamic stress on the blade during rapid vibration.

Claims

1. A method for automatic and rapid processing of dynamic stress test data, characterized in that: Includes the following steps: ① Distinguish signal types: Based on the changes in the original signal, determine the type of test section the current data belongs to; the original signal includes rotational speed data and vibration data; ② Adjust time and frequency resolution: Set the time resolution and frequency resolution according to the type of test section, and adjust the time resolution... and frequency resolution Meet the conditions ; ③ Transformation calculation: The original signal is processed using the short-time Fourier transform; In step ①, the following steps are specifically adopted: S1.

1. Denoising: Denoise the entire data segment of length L; S1.

2. Piecewise Fitting: A linear function is used to fit each segment of data to obtain the fitting equation for each segment; S1.

3. Calculate the mean: Calculate the average and root mean square of the slope of each data segment; S1.

4. Definition and Classification: Based on the average value and root mean square, the slope of each data segment is classified into types greater than the threshold and types less than the threshold. Types greater than the threshold include acceleration and deceleration states, while types less than the threshold include steady states. S1.

5. Specific classification: Differential equations are used to distinguish the data segments with values ​​greater than the threshold and the type of test segment to which they belong; In step S1.4, the threshold is = + ,in, This represents the average slope of the data segments. This represents the root mean square of the slope of each data segment.

2. The automatic and rapid processing method for dynamic stress test data as described in claim 1, characterized in that: The short-time Fourier transform formula is as follows: In the formula, For window functions, The original signal is represented by *, and the conjugate signal is represented by *. The center of the window function, Regarding time and frequency A two-dimensional function.

3. The automatic and rapid processing method for dynamic stress test data as described in claim 2, characterized in that: The short-time Fourier transform is obtained by discretizing the time-frequency grid (mT, nF) with equal intervals using the following formula: In the formula, All values ​​are integers, representing different engine states automatically determined based on engine speed. , Represent the different time resolutions and frequency resolutions used in each state, and satisfy the following conditions: >0, >0, It is a two-dimensional function of mT and nF.

4. The automatic and rapid processing method for dynamic stress test data as described in claim 1, characterized in that: In step S1.1, it is divided into n segments, n = L / L win Let each data segment be N1, N2, ..., N n .

5. The automatic and rapid processing method for dynamic stress test data as described in claim 1, characterized in that: In step S1.2, the fitting equation for each segment is Y. 0i =k 0i +b 0i (i=1,2,…,n), if the coefficient of the linear term k 0i If the value is greater than or equal to zero, then the data segment is sorted in ascending order; if the coefficient of the linear term k is greater than or equal to zero... 0i If the value is less than zero, the data segment is sorted in descending order.

6. The automatic and rapid processing method for dynamic stress test data as described in claim 5, characterized in that: The data after being rearranged is fitted again using a linear function, and the fitting equation for each data segment is Y1i=k1i+b1i (i=1,2,…,n), where k1i represents the rate of change of rotation speed with time, i.e., the slope.

7. The automatic and rapid processing method for dynamic stress test data as described in claim 1, characterized in that: In step S1.3, the average slope of each data segment is: ； The root mean square of the slope of each data segment is: 。 8. The automatic and rapid processing method for dynamic stress test data as described in claim 1, characterized in that: In step S1.5, the difference equation is Y=[x(2)-x(1),x(3)-x(2),…, x(n)-x(n-1)], the length of Y is n-1 and it is stored in array form. In array Y, -1 indicates that the process is the beginning of the deceleration state and 1 indicates that the process is the beginning of the acceleration state.

Images