Efficient Compression and Editing Method for Automotive Component Load Spectra Based on Wigner-Ville Transform

Abstract

An efficient compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform includes the following steps: 1) Preprocess the original signal; 2) Perform Hilbert transform on the preprocessed signal to obtain the analytic signal; 3) Perform Wigner-Ville transform on the analytic signal to obtain the Wigner-Ville distribution; 4) Integrate the Wigner-Ville distribution to obtain the instantaneous energy spectrum; 5) Use the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum to obtain the retained signal segments; 6) Stitch the retained signal segments to obtain the compressed load signal; 7) Determine whether the relative damage retention and the error of the statistical parameters meet the threshold requirements, and judge whether the power spectral density and the distribution of the level-crossing count curve of the load signal before and after compression meet the requirements. If not, return to step 5) until the requirements are met; 8) Output the compressed load signal. The present invention can accurately identify and extract the part with high fatigue damage contribution, quickly compress and edit to form an accelerated load spectrum, and achieve the same test effect as the original signal.

Description

Technical Field

[0001] The invention relates to the field of indoor road simulation acceleration test of automobile parts, in particular to a high-efficiency compression editing method of automobile parts load spectrum based on Wigner-Ville transformation. Background Art

[0002] Durability analysis of automotive parts is an important part of the product development process of automobile companies, and the load information acting on parts is the key to analyzing fatigue durability issues. For the measured road load spectrum loading, a reasonable and effective load spectrum compilation method is adopted to accelerate editing. On the basis of ensuring the consistency of the load spectrum loading effect, a load spectrum with a shorter loading time is obtained to significantly improve the research efficiency of part fatigue durability testing. In order to ensure that the reduced load spectrum has the same loading effect on the parts as the original load spectrum, it is necessary to make the reduced load spectrum basically consistent with the original signal in terms of damage contribution, statistical parameters (mean, root mean square value and peak factor), power spectrum density and penetration count.

[0003] At present, load spectrum accelerated editing is mainly divided into two categories: time domain editing and frequency domain editing. The basic principle is to shorten the data length by deleting the load cycles that do not contribute much to the damage in the signal, and ensure that the damage contribution to the parts is basically consistent with the original signal. The differences between the editing methods are mainly reflected in the identification and deletion of small damage contribution data, including setting indicators such as strain amplitude, SWT threshold, and damage retention to identify and eliminate invalid signal segments to achieve load spectrum compression editing.

[0004] The existing technology for accelerating the compilation of load spectra of automotive parts includes: using fatigue analysis software such as Ncode, setting the damage retention amount, and compressing the load spectrum based on time domain damage retention. However, the compression effect of this method is limited, and it is easy to cause the compressed signal to change significantly in statistical characteristic parameters, frequency domain and amplitude domain distribution compared with the original signal, and the difference is too large. Summary of the invention

[0005] The purpose of the present invention is to propose a method for compressing and editing the load spectrum of automobile parts based on Wigner-Ville transformation. The method uses Wigner-Ville time-frequency processing technology to jointly analyze the time-frequency domain characteristics of the original load signal, and can obtain the instantaneous energy information corresponding to each time point of the load signal. According to the instantaneous energy spectrum obtained, combined with the genetic algorithm to optimize the optimal threshold, the invalid damage contribution data of the original load signal can be clearly identified and deleted, and the original load signal time length can be compressed to the maximum extent, and the compressed signal is basically consistent with the original signal in terms of damage retention, statistical parameters (mean, root mean square value and peak coefficient), power spectrum density distribution and level-crossing count, so as to achieve a loading effect consistent with the original signal.

[0006] The technical solution adopted to achieve the purpose of the present invention is as follows. An efficient compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform includes the following steps:

[0007] 1) Obtain the original load signal.

[0008] 2) Preprocess the original load signal to obtain the preprocessed load signal.

[0009] 3) Perform the Hilbert transform on the preprocessed load signal to obtain the analytic signal.

[0010] 4) Perform the Wigner-Ville transform on the analytic signal to obtain the information of the energy distribution over time-frequency, and obtain the two-dimensional time-frequency real matrix and the Wigner-Ville spectrum.

[0011] 5) At each time point of the two-dimensional time-frequency real matrix, integrate the Wigner-Ville spectrum along the frequency axis to obtain the instantaneous energy spectrum of the original load signal.

[0012] 6) Set the parameters of the genetic algorithm, use the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum, and according to the optimal threshold, locate the segments of the instantaneous energy spectrum below the threshold and the time points of the corresponding data, map the time points to the time series of the original load signal, and delete the segments of the original load signal corresponding to these time points to obtain the retained signal segments.

[0013] 7) Stitch the retained signal segments to obtain the compressed load signal.

[0014] 8) Calculate the relative damage retention and the error of the statistical parameters between the compressed load signal and the original load signal, and determine the frequency-domain power spectral density and the amplitude-domain level-crossing count curve distribution of the compressed load signal and the original load signal.

[0015] 9) Determine whether the relative damage retention and the error of the statistical parameters meet the threshold requirements.

[0016] Judge whether the main distributions of the frequency-domain power spectral density of the load signals before and after compression are consistent.

[0017] Judge whether the amplitude-domain level-crossing count distributions of the load signals before and after compression are consistent.

[0018] If the relative damage retention is lower than the threshold, the error of the statistical parameters exceeds the threshold, the frequency-domain power spectral density of the load signals before and after compression is not consistent, or the trend of the amplitude-domain level-crossing count distribution curves of the load signals before and after compression is not consistent, then return to step 6), otherwise, go to step 10).

[0019] 10) Output the compressed load signal.

[0020] Furthermore, the preprocessing of the original load signal includes: filtering, deburring, drift correction, and resampling.

[0021] Furthermore, the analytic signal z(t) after Hilbert transform is as follows:

[0022] z(t)＝x(t)+jH[x(t)] (1)

[0023] where t represents the time vector, x(t) is the original load signal, and z(t) is the analytic signal of x(t).

[0024] The Hilbert transform H[x(t)] is as follows:

[0025]

[0026] where PV represents the Cauchy principal value, and τ is the integration variable.

[0027] Furthermore, the Wigner-Ville distribution W z (nT, ω) after Wigner-Ville transform is as follows:

[0028]

[0029] where W z (nT, ω) is the Wigner-Ville distribution, nT and ω represent the indices of the time vector and frequency vector respectively. L is the length of the time data used for the Wigner-Ville transform, z * (t) is the complex conjugate of the analytic signal z(t), and z(nT + iT)z * (nT - iT) is the instantaneous correlation function of the analytic signal z(t). i represents the time data sequence number.

[0030] The two-dimensional time-frequency real matrix is as follows:

[0031]

[0032] where TFR(t, f) is the two-dimensional time-frequency real matrix, t is the time vector, f is the frequency, and the row vector a m = [a m1 , a m2 ,..., a mn corresponds to different time points, and the column vector a n = [a 1n , a 2n ,..., a mn TRepresent different frequency values. m and n are the number of rows and columns. The elements of the two-dimensional time-frequency real matrix represent the amplitude and phase angle information of the signal.

[0033] Further, the instantaneous energy spectrum of the original load signal is as follows:

[0034]

[0035] In the formula, |Z(t)| 2 is the instantaneous energy spectrum. W z (t, f) is the Wigner-Ville spectrum.

[0036] Further, the steps of using the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum include:

[0037] Taking the threshold in the instantaneous energy spectrum as the design variable, taking the minimum value min(L(y)) of the length compression ratio before and after load compression and editing as the objective function, and taking the damage ratio between the compressed signal and the original signal not less than a and the statistical parameter error less than b as the constraint conditions, establish a threshold optimization mathematical model. a and b are preset error thresholds.

[0038] Use the threshold optimization mathematical model to perform threshold optimization calculation to find the optimal threshold within the numerical range [0, max P(n)] of the instantaneous energy spectrum.

[0039] Among them, the threshold optimization mathematical model is as follows:

[0040]

[0041] In the formula, Δavr, Δku, and Δrms are the mean value, kurtosis coefficient, and root mean square value change amount respectively. L(y) is the length compression ratio before and after load compression and editing. L y is the length after load compression and editing, and L0 is the length before load compression and editing. is the relative damage retention amount of the signal before and after editing. d(x0(t)) and d(x y (t)) are the pseudo-damage values of the signal before and after editing respectively.

[0042] Among them, the pseudo-damage d(x0(t)) and d(x y (t)) of the signal before and after editing are theoretical values, without average stress correction. The damage values of each stress cycle calculated are as follows for the total pseudo-damage calculation of the load data:

[0043]

[0044] In the formula, D represents the total pseudo-damage, n j is the number of cycles of the jth stress level. N jis the number of cycles to failure at this stress level.

[0045] Further, in step 7), when there is no intersection in the time series corresponding to the instantaneous energy spectrum segment in the original signal, the remaining segments in the original signal can be directly spliced to obtain the compressed signal.

[0046] Further, the statistical parameters include: mean value, root mean square value, and kurtosis coefficient.

[0047] Further, the frequency-domain characteristics of the load signal are represented as a power spectral density curve, and the amplitude-domain characteristics are represented as a cross-level count distribution.

[0048] The technical effect of the present invention is beyond doubt. The load spectrum compression and editing method for automotive parts based on the Wigner-Ville transform of the present invention has the following beneficial effects:

[0049] The load spectrum compression and editing method for automotive parts based on the Wigner-Ville transform can jointly analyze the time-frequency domain distribution for random non-stationary signals, and obtain the energy information corresponding to each time point, which is fundamentally different from the traditional editing method based on time-domain damage retention. It can not only ensure the consistency of characteristic parameters before and after signal compression in the time domain, but also maintain consistency in the frequency-domain and amplitude-domain characteristic distributions;

[0050] The load spectrum compression and editing method for automotive parts based on the Wigner-Ville transform has a better time compression effect for the signal of the present invention compared with the editing methods based on time-domain damage retention, short-time Fourier transform, and S-transform under the same damage retention amount;

[0051] The load spectrum compression and editing method for automotive parts based on the Wigner-Ville transform can compress more time data and have a smaller data volume within reasonable constraints compared with the editing method based on time-domain damage retention, and can more effectively improve the efficiency of fatigue simulation tests or bench tests;

[0052] The load spectrum compression and editing method for automotive parts based on the Wigner-Ville transform has the same trend as the power spectral density distribution of the original load in the low-frequency region for the power spectral density curve in the frequency domain of the present invention compared with the editing method based on time-domain damage retention. Due to the deletion of more small load amplitudes in the present invention, the power spectral density distribution curve shifts more upward in the high-frequency region than the latter;

[0053] The method for compressing and editing the load spectrum of automotive parts based on the Wigner-Ville transform, compared with the editing method based on damage retention in the time domain, has the same distribution trend of the cross-level counting curve in the amplitude domain and the cross-level counting curve of the original load. The present invention deletes more small-amplitude cycles near the mean value, that is, the compression effect is better;

[0054] The method for compressing and editing the load spectrum of automotive parts based on the Wigner-Ville transform, compared with the editing methods based on the short-time Fourier transform and the wavelet transform, simplifies the editing operation while ensuring data accuracy, and avoids steps such as the selection and setting of window functions or wavelet functions;

[0055] Based on the time-frequency analysis method of the Wigner-Ville distribution, the present invention can accurately identify and extract the data part with high fatigue damage contribution, quickly compress and form an accelerated load spectrum, and compress the original load signal to the greatest extent while ensuring that the obtained compressed signal is basically consistent with the original signal, achieving the same loading effect as the original signal. Brief Description of the Drawings

[0056] Figure 1 is the flowchart of the method of the present invention;

[0057] Figure 2 is the Wigner-Ville spectrogram showing the distribution of signal energy in the time-frequency domain;

[0058] Figure 3 is the instantaneous energy spectrum of the load signal;

[0059] Figure 4 is a schematic diagram of extracting the signal time segment based on the threshold setting of the instantaneous energy spectrum;

[0060] Figure 5 is a schematic diagram of extracting the time-domain signal segment with a large damage contribution;

[0061] Figure 6 is the accelerated spectrum after editing the load signal. Detailed Embodiments

[0062] The present invention will be further described below in conjunction with embodiments, but it should not be understood that the above-mentioned subject matter scope of the present invention is limited to the following embodiments. Without departing from the above-mentioned technical ideas of the present invention, various substitutions and changes made according to common general technical knowledge and customary means in the art should be included within the protection scope of the present invention.

[0063] Embodiment 1:

[0064] See Figures 1 to 6 , the method for efficiently compressing and editing the load spectrum of automotive parts based on the Wigner-Ville transform includes the following steps:

[0065] 1) Obtain the original load signal. The original load signal is the measured load signal.

[0066] 2) Preprocess the original load signal to obtain the preprocessed load signal.

[0067] 3) Perform Hilbert transform on the preprocessed load signal to obtain the analytic signal.

[0068] 4) Perform Wigner-Ville transform on the analytic signal to obtain the information of energy distribution over time-frequency, and obtain the two-dimensional time-frequency real matrix and the Wigner-Ville spectrum.

[0069] The Wigner-Ville transform is defined as a function of energy distribution over time and its frequency.

[0070] 5) At each time point of the two-dimensional time-frequency real matrix, integrate the Wigner-Ville spectrum along the frequency axis to obtain the instantaneous energy spectrum of the original load signal.

[0071] 6) Set the genetic algorithm parameters, use the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum, and according to the optimal threshold, locate the instantaneous energy spectrum segments below the threshold and the time points of the corresponding data, map the time points to the time series of the original load signal, and delete the original load signal segments corresponding to these time points. These deleted original load signal segments are the data segments contributed by small damages. After deletion, it does not affect the accuracy of the load signal, and the remaining signal segments are obtained.

[0072] 7) Stitch the remaining signal segments to obtain the compressed load signal.

[0073] 8) Calculate the relative damage retention and the error of statistical parameters between the compressed load signal and the original load signal, and determine the frequency-domain power spectral density and the amplitude-domain level-crossing count curve distribution of the compressed load signal and the original load signal.

[0074] 9) Determine whether the relative damage retention and the error of statistical parameters meet the threshold requirements.

[0075] Judge whether the main distributions of the frequency-domain power spectral density of the load signal before and after compression are consistent.

[0076] The method for judging whether there is consistency is: judge whether the distribution has a significantly larger difference, that is, judge whether the distribution difference is greater than the preset threshold.

[0077] Judge whether the amplitude-domain level-crossing count distributions of the load signal before and after compression are consistent.

[0078] If the relative damage retention amount is lower than the threshold value, the error of the statistical parameter exceeds the threshold value, the frequency-domain power spectral density of the load signal before and after compression does not have consistency, or the trend of the amplitude-domain level crossing count distribution curve of the load signal before and after compression does not have consistency, then return to step 6); otherwise, proceed to step 10).

[0079] 10) Output the compressed load signal. The loading time of the compressed load signal is short, and this signal is used to analyze the fatigue durability of parts.

[0080] Through this signal, the fatigue durability test of parts can be accelerated and the test efficiency can be improved.

[0081] The preprocessing performed on the original load signal includes: filtering, deburring, drift correction, and resampling processing.

[0082] The Hilbert transform is used to eliminate the influence of negative frequency components in the real signal and avoid frequency-domain aliasing. The analytic signal z(t) after the Hilbert transform is as follows:[[]]

[0083] z(t) = x(t) + jH[x(t)] (1)

[0084] In the formula, t represents the time vector, x(t) is the original load signal, and z(t) is the analytic signal of x(t).

[0085] The Hilbert transform H[x(t)] is as follows:[[]]

[0086]

[0087] In the formula, PV represents the Cauchy principal value, and τ is the integration variable.

[0088] The Wigner-Ville distribution W z (nT, ω) after the Wigner-Ville transform is as follows:[[]]

[0089]

[0090] In the formula, W z (nT, ω) is the Wigner-Ville distribution, and nT and ω respectively represent the indices of the time vector and the frequency vector; L is the length of the time data used for the Wigner-Ville transform, and z * (t) is the complex conjugate of the analytic signal z(t), and z(nT + iT)z * (nT - iT) is the instantaneous correlation function of the analytic signal z(t); i represents the time data sequence number.

[0091] Performing the Wigner-Ville transform on the discrete signal can obtain the three-dimensional distribution information of time-frequency-energy of the signal.

[0092] The two-dimensional time-frequency real matrix is as follows:

[0093]

[0094] In the formula, TFR(t, f) is the two-dimensional time-frequency real matrix, t is the time vector, f is the frequency, and the row vector a m = [a m1 , a m2 ,..., a mn corresponds to different time points, and the column vector a n = [a 1n , a 2n ,..., a mn T represents different frequency values. m and n are the number of rows and columns. The elements of the two-dimensional time-frequency real matrix represent the amplitude and phase angle information of the signal.

[0095] The Wigner-Ville distribution has the time marginal property. From the two-dimensional time-frequency real matrix, at each time point, the integral along the frequency domain axis is equal to the instantaneous energy of the signal at the corresponding moment. Then, the energy corresponding to each time point of the load signal is calculated to form the instantaneous energy spectrum. The integration of the Wigner-Ville spectrum is performed along the frequency axis direction at each time point to obtain the instantaneous energy corresponding to each time point of the signal. The cut-off threshold is closely related to the time history length of the compressed load spectrum signal and the retained amount of fatigue damage.

[0096] The instantaneous energy spectrum of the original load signal is as follows:

[0097]

[0098] In the formula, |Z(t)| 2 is the instantaneous energy spectrum; W z (t, f) is the Wigner-Ville spectrum.

[0099] The steps of using the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum include:

[0100] Taking the threshold in the instantaneous energy spectrum as the design variable, taking the minimum value min(L(y)) of the length compression ratio before and after load compression editing as the objective function, and taking the damage ratio between the compressed signal and the original signal not less than a and the statistical parameter error less than b as the constraint conditions, a threshold optimization mathematical model is established. a and b are preset error thresholds.

[0101] Using the threshold optimization mathematical model to perform threshold optimization calculation to find the optimal threshold within the numerical range [0, max P(n)] of the instantaneous energy spectrum. ​

[0102] Among them, the threshold optimization mathematical model is as follows:

[0103]

[0104] In the formula, Δavr, Δku, and Δrms are the change amounts of the mean value, kurtosis coefficient, and root mean square value respectively. L(y) is the length compression ratio before and after the load compression editing. L y is the length after the load compression editing, and L0 is the length before the load compression editing. is the relative damage retention amount of the signal before and after editing; d(x0(t)) and d(x y (t)) are the pseudo-damage values of the signal before and after editing respectively;

[0105] Among them, the pseudo-damage d(x0(t)) and d(x y (t)) of the signal before and after editing are theoretical values, which is a method to reliably reflect the fatigue characteristics of the load spectrum with simple numerical values. It is not real damage and is usually used to compare the relationships between different load histories. It is mainly used in the field of structural durability tests, especially in the evaluation of the editing effect during the bench drive spectrum editing stage. Without average stress correction, the damage values of each stress cycle are calculated, and the total pseudo-damage of the load data is calculated as follows:

[0106]

[0107] In the formula, D represents the total pseudo-damage, and n j is the number of cycles of the j-th stress level. N j is the number of cycles to failure at this stress level.

[0108] In step 7), when there is no intersection in the time series corresponding to the instantaneous energy spectrum segment in the original signal, the retained segments in the original signal can be directly spliced to obtain the compressed signal.

[0109] The statistical parameters include: mean value, root mean square value, and kurtosis coefficient.

[0110] The frequency domain characteristics of the load signal are represented as a power spectral density curve, and the amplitude domain characteristics are represented as a cross-level count distribution.

[0111] Embodiment 2:

[0112] Refer to Figures 1 to 6 , the efficient compression editing method for the load spectrum of automotive parts based on the Wigner-Ville transform includes the following steps:

[0113] 1) Obtain the original load signal. The original load signal is the measured load signal.

[0114] 2) Preprocess the original load signal to obtain the preprocessed load signal.

[0115] 3) Perform Hilbert transform on the preprocessed load signal to obtain the analytic signal.

[0116] 4) Perform Wigner-Ville transform on the analytic signal to obtain the information of energy distribution over time-frequency, and obtain the two-dimensional time-frequency real matrix and the Wigner-Ville spectrum.

[0117] The Wigner-Ville transform is defined as a function of energy distribution over time and its frequency.

[0118] 5) At each time point of the two-dimensional time-frequency real matrix, integrate the Wigner-Ville spectrum along the frequency axis to obtain the instantaneous energy spectrum of the original load signal.

[0119] 6) Set the genetic algorithm parameters, use the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum, and according to the optimal threshold, locate the segments of the instantaneous energy spectrum below the threshold and the time points of the corresponding data, map the time points to the time series of the original load signal, delete the segments of the original load signal corresponding to these time points. These deleted segments of the original load signal are the data segments contributed by small damages. After deletion, it does not affect the accuracy of the load signal, and the remaining signal segments are obtained.

[0120] 7) Stitch the remaining signal segments to obtain the compressed load signal.

[0121] 8) Calculate the relative damage retention and the error of statistical parameters between the compressed load signal and the original load signal, and determine the frequency-domain power spectral density and the amplitude-domain level crossing count curve distribution of the compressed load signal and the original load signal.

[0122] 9) Determine whether the relative damage retention and the error of statistical parameters meet the threshold requirements.

[0123] Judge whether the main distributions of the frequency-domain power spectral density of the load signal before and after compression are consistent.

[0124] The method for judging whether there is consistency is: judge whether the distribution has a significantly large difference, that is, judge whether the distribution difference is greater than the preset threshold.

[0125] Judge whether the amplitude-domain level crossing count distributions of the load signal before and after compression are consistent.

[0126] If the relative damage retention is lower than the threshold, the error of statistical parameters exceeds the threshold, the frequency-domain power spectral density of the load signal before and after compression is inconsistent, or the trend of the amplitude-domain level crossing count distribution curve of the load signal before and after compression is inconsistent, then return to step 6); otherwise, go to step 10).

[0127] 10) Output a compressed load signal. The loading time of the compressed load signal is short, and this signal is used to analyze the fatigue durability of parts.

[0128] Through this signal, the fatigue durability test of parts can be accelerated and the test efficiency can be improved.

[0129] Embodiment 3:

[0130] For the efficient compression and editing method of the load spectrum of automotive parts based on the Wigner-Ville transform, the main steps are shown in Embodiment 2. The preprocessing performed on the original load signal includes: filtering, deburring, drift correction, and resampling processing.

[0131] Embodiment 4:

[0132] For the efficient compression and editing method of the load spectrum of automotive parts based on the Wigner-Ville transform, the main steps are shown in Embodiment 2. The Hilbert transform is used to eliminate the influence of negative frequency components in the real signal and avoid frequency domain aliasing. The analytic signal z(t) after the Hilbert transform is as follows:

[0133] z(t) = x(t) + jH[x(t)] (1)

[0134] In the formula, t represents the time vector, x(t) is the original load signal, and z(t) is the analytic signal of x(t).

[0135] The Hilbert transform H[x(t)] is as follows:

[0136]

[0137] In the formula, PV represents the Cauchy principal value, and τ is the integration variable.

[0138] Embodiment 5:

[0139] For the efficient compression and editing method of the load spectrum of automotive parts based on the Wigner-Ville transform, the main steps are shown in Embodiment 2. The Wigner-Ville distribution W z (nT, ω) after the Wigner-Ville transform is as follows:

[0140]

[0141] In the formula, W z (nT, ω) is the Wigner-Ville distribution, and nT, ω represent the indices of the time vector and the frequency vector respectively; L is the length of the time data used for the Wigner-Ville transform, z *(t) is the complex conjugate of the analytic signal z(t), and z(nT + iT)z * (nT - iT) is the instantaneous correlation function of the analytic signal z(t); i represents the time data sequence number.

[0142] Performing the Wigner-Ville transform on the discrete signal can obtain the three-dimensional distribution information of time-frequency-energy of the signal.

[0143] The two-dimensional time-frequency real matrix is shown as follows:

[0144]

[0145] In the formula, TFR(t, f) is the two-dimensional time-frequency real matrix, t is the time vector, f is the frequency, and the row vector a m = [a m1 , a m2 ,..., a mn corresponds to different time points, and the column vector a n = [a 1n , a 2n ,..., a mn T represents different frequency values. m and n are the number of rows and columns. The elements of the two-dimensional time-frequency real matrix represent the amplitude and phase angle information of the signal.

[0146] Example 6:

[0147] For the efficient compression and editing method of the load spectrum of automotive parts based on the Wigner-Ville transform, the main steps are shown in Example 2. The Wigner-Ville distribution has the time marginal property. From the two-dimensional time-frequency real matrix, at each time point, the integral along the frequency domain axis is equal to the instantaneous energy of the signal at the corresponding moment. Then, the energy corresponding to each time point of the load signal is calculated to form the instantaneous energy spectrum. The integration of the Wigner-Ville spectrum is performed at each time point along the frequency axis direction to obtain the instantaneous energy corresponding to each time point of the signal. The cut-off threshold is closely related to the time history length of the compressed load spectrum signal and the retained amount of fatigue damage.

[0148] The instantaneous energy spectrum of the original load signal is shown as follows:

[0149]

[0150] In the formula, |Z(t)| 2 is the instantaneous energy spectrum; W z (t, f) is the Wigner-Ville spectrum.

[0151] Example 7:

[0152] ​An efficient compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform. The main steps are shown in Example 2. The steps of using the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum include:

[0153] Taking the threshold in the instantaneous energy spectrum as the design variable, taking the minimum value min(L(y)) of the length compression ratio before and after load compression and editing as the objective function, and taking the damage ratio between the compressed signal and the original signal not less than 97% and the statistical parameter error less than 15% as the constraint conditions, a threshold optimization mathematical model is established.

[0154] Using the threshold optimization mathematical model to perform threshold optimization calculation to find the optimal threshold within the numerical range [0, max P(n)] of the instantaneous energy spectrum.

[0155] Among them, the threshold optimization mathematical model is as follows:

[0156]

[0157] In the formula, Δavr, Δku, and Δrms are the changes in the mean value, kurtosis coefficient, and root mean square value respectively. L(y) is the length compression ratio before and after load compression and editing. L y is the length after load compression and editing, and L0 is the length before load compression and editing. is the relative damage retention amount of the signal before and after editing; d(x0(t)) and d(x y (t)) are the pseudo-damage values of the signal before and after editing respectively;

[0158] Among them, the pseudo-damage d(x0(t)) and d(x y (t)) of the signal before and after editing are theoretical values. It is a method that can reliably reflect the fatigue characteristics of the load spectrum with simple numerical values. It is not real damage and is usually used to compare the relationships between different load histories. It is mainly used in the field of structural durability tests, especially widely used in the evaluation of the editing effect during the editing stage of the bench drive spectrum. Without average stress correction, the damage values of each stress cycle are calculated, and the total pseudo-damage of the load data is calculated as follows:

[0159]

[0160] In the formula, D represents the total pseudo-damage, n j is the number of cycles of the jth stress level. N j is the number of cycles to failure at this stress level.

[0161] Example 8:

[0162] An efficient compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform. The main steps are shown in Embodiment 2. In step 7), when there is no intersection in the time series corresponding to the instantaneous energy spectrum segment in the original signal, the retained segments in the original signal can be directly spliced to obtain the compressed signal.

[0163] Embodiment 9:

[0164] An efficient compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform. The main steps are shown in Embodiment 2. The statistical parameters include: mean value, root mean square value, and kurtosis coefficient.

[0165] Embodiment 10:

[0166] An efficient compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform. The main steps are shown in Embodiment 2. The frequency-domain characteristics of the load signal are represented by the power spectral density curve, and the amplitude-domain characteristics are represented by the through-level count distribution.

[0167] Embodiment 11:

[0168] See Figures 1 to 6 , an efficient compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform, which specifically includes the following steps:

[0169] 1) Obtain and input the original load signal;

[0170] 2) Preprocess the load spectrum, including filtering, deburring, drift correction, and resampling;

[0171] 3) Perform Hilbert transform on the preprocessed signal to obtain the analytical signal form of the real signal;

[0172] 4) Perform Wigner-Ville transform on the analytical signal to obtain the information of energy distribution over time-frequency, and obtain a two-dimensional time-frequency real matrix and the Wigner-Ville spectrum;

[0173] 5) For the Wigner-Ville spectrum, integrate along the frequency axis at each time point to obtain the instantaneous energy corresponding to each time point of the signal, and obtain the instantaneous energy spectrum of the load signal;

[0174] 6) Combine genetic algorithm to optimize the threshold selection. For the instantaneous energy spectrum, use genetic algorithm to determine the optimal threshold. According to the obtained threshold, locate the instantaneous energy spectrum segments below the threshold, and locate the corresponding time points of the data. Map the time points to the original load time series, and delete the corresponding data segments with small damage contributions;

[0175] 7) Stitch the reserved signal segments to obtain the compressed signal;

[0176] 8) Calculate the relative damage retention and statistical parameter error between the compressed load signal and the original load signal, test the power spectral density and the distribution of the level crossing count curve, determine whether the damage retention and statistical parameter error meet the requirements, and whether the power spectral density in the frequency domain of the signal before and after compression is reasonable, and whether the trend of the level crossing count distribution curve in the amplitude domain is consistent. If there are obvious differences, loop back to step 6) to reset the genetic algorithm parameters and obtain the optimal threshold until the engineering requirements are met, then jump out of the loop;

[0177] 9) Complete the compression and editing work of the load spectrum of automotive parts.

[0178] The Hilbert transform in step 3) is shown in formula (1), which is used to eliminate the influence of negative frequency components in the real signal and avoid frequency domain aliasing.

[0179]

[0180] Among them, x(t) is the original load signal, z(t) is the analytic signal form of x(t), H[x(t)] is the Hilbert transform, and PV represents the Cauchy principal value.

[0181] The Wigner-Ville transform in step 4) includes: The Wigner-Ville transform is defined as a function of energy distribution over time and its frequency, as shown in formula (2), and the discrete Wigner-Ville transform form of the analytic signal is shown in formula (3). Performing the Wigner-Ville transform on the discrete signal can obtain the three-dimensional distribution information of time-frequency-energy of the signal. Its two-dimensional time-frequency real number matrix is shown in formula (4), and the row vector a m =[a m1 ,a m2 ,...,a mn corresponds to different time points, and the column vector a n =[a 1n ,a 2n ,...,a mn T represents different frequency values, and the matrix elements represent the amplitude and phase angle information of the signal.

[0182]

[0183]

[0184]

[0185] Among them, f is the frequency, τ is the integration variable, z * ​(t) is the complex conjugate of z(t). is the instantaneous correlation function of the signal. W z W(t, ω) represents the Wigner-Ville distribution, T is the sampling period, L is the length of the time data used for the transformation, and t and ω represent the indices of the time vector and the frequency vector respectively. TFR(t, f) is a two-dimensional time-frequency real-valued matrix.

[0186] The method for obtaining the instantaneous energy spectrum of the load signal in step 5) is as follows: The Wigner-Ville distribution has the time marginal property. From the two-dimensional time-frequency real matrix, at each time point, the integral along the frequency domain axis is equal to the instantaneous energy of the signal at the corresponding moment, as shown in formula (5). Then, the energy corresponding to the load signal at each time point is calculated to form the instantaneous energy spectrum.

[0187]

[0188] The steps of setting the threshold for the instantaneous energy spectrum in step 6) specifically include: Combining the engineering requirements, using the genetic algorithm, taking the instantaneous energy spectrum threshold as the design variable, taking the minimum min(L(y)) of the length compression ratio before and after load editing as the objective function, and taking the damage ratio between the compressed signal and the original signal not less than 97% and the statistical parameter error less than 15% as the constraint conditions, performing threshold optimization calculation to find the optimal threshold within the numerical range [0, max P(n)] of the instantaneous energy spectrum. The cut-off threshold is closely related to the time history length of the compressed load spectrum signal and the remaining amount of fatigue damage. The optimization mathematical model of the threshold is expressed as shown in formula (6).

[0189]

[0190] Among them, is the relative damage retention amount of the signal before and after editing, and Δavr, Δku, and Δrms are the change amounts of the mean value, kurtosis coefficient, and root mean square value respectively. The calculation of the pseudo-damage value is shown in formula (7).

[0191]

[0192] Among them, D represents the total pseudo-damage, and n i is the number of cycles at a certain stress level; N i is the number of cycles to failure at this stress level.

[0193] The steps of splicing the retained signal segments in step 7) specifically include: There is no intersection in the time series corresponding to the instantaneous energy spectrum segments in the original signal, so the retained segments in the original signal can be directly spliced to obtain the compressed signal.

[0194] The statistical parameters in step 8) include: mean value, root mean square value, and kurtosis coefficient; the frequency domain characteristics are represented as a power spectral density curve, and the amplitude domain characteristics are represented as a cross-level count distribution.

Claims

1. An efficient compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform, characterized in that, Including the following steps: 1) Obtain the original load signal; 2) Preprocess the original load signal to obtain the preprocessed load signal; 3) Perform Hilbert transform on the preprocessed load signal to obtain the analytic signal; 4) Perform Wigner-Ville transform on the analytic signal to obtain the information of energy distribution over time-frequency, and obtain the two-dimensional time-frequency real matrix and the Wigner-Ville spectrum; 5) At each time point of the two-dimensional time-frequency real matrix, integrate the Wigner-Ville spectrum along the frequency axis to obtain the instantaneous energy spectrum of the original load signal; 6) Set the genetic algorithm parameters, use the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum, and according to the optimal threshold, locate the segments of the instantaneous energy spectrum below the threshold and the time points of the corresponding data, map the time points to the time series of the original load signal, and delete the segments of the original load signal corresponding to these time points to obtain the retained signal segments; 7) Stitch the retained signal segments to obtain the compressed load signal; 8) Calculate the relative damage retention and the error of statistical parameters between the compressed load signal and the original load signal, and determine the frequency-domain power spectral density and the amplitude-domain level-crossing count curve distribution of the compressed load signal and the original load signal; 9) Determine whether the relative damage retention and the error of statistical parameters meet the threshold requirements; Judge whether the main distributions of the frequency-domain power spectral density of the load signals before and after compression are consistent; Judge whether the amplitude-domain level-crossing count distributions of the load signals before and after compression are consistent; If the relative damage retention is lower than the threshold, the error of the statistical parameters exceeds the threshold, the frequency-domain power spectral density of the load signals before and after compression is inconsistent, or the trend of the amplitude-domain level-crossing count distribution curve of the load signals before and after compression is inconsistent, then return to step 6), otherwise, go to step 10); 10) Output the compressed load signal.

2. The high-efficiency compression and editing method of the vehicle component load spectrum based on the Wigner-Ville transform according to claim 1, characterized in that The preprocessing performed on the original load signal includes: filtering, deburring, drift correction, and resampling processing.

3. The high-efficiency compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform according to claim 1, wherein The analytic signal z(t) after the Hilbert transform is as follows: z(t) = x(t) + jH[x(t)] (1) In the formula, t represents the time vector, x(t) is the original load signal, and z(t) is the analytic signal of x(t); The Hilbert transform H[x(t)] is as follows: In the formula, PV represents the Cauchy principal value, and τ is the integration variable.

4. The high-efficiency compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform according to claim 1, characterized in that The Wigner-Ville distribution $W$ after the Wigner-Ville transform z (nT, ω) is as follows: where W z (nT, ω) is the Wigner-Ville distribution, and T and ω respectively represent the indices of the time vector and the frequency vector; L is the length of the time data for the Wigner-Ville transform, z * (t) is the complex conjugate of the analytic signal z(t), and z(nT + iT)z * (nT - iT) is the instantaneous correlation function of the analytic signal z(t); i represents the sequence number of the time data; The two-dimensional time-frequency real matrix is as follows: where TFR(t,f) is a two-dimensional time-frequency real matrix, t is the time vector, f is the frequency, and the row vector a m = [a m1 , a m2 , …, a mn corresponds to different time points, and the column vector a n = [a 1n , a 2n , …, a mn T represents different frequency values; m and n are the number of rows and columns; the elements of the two-dimensional time-frequency real matrix represent the amplitude and phase angle information of the signal.​ 5. The high-efficiency compression and editing method for the load spectrum of automotive parts based on the Wigner-Ville transform according to claim 1, wherein The instantaneous energy spectrum of the original load signal is as follows: where, |Z(t) 2 is the instantaneous energy spectrum; W z (t, f) is the Wigner-Ville spectrum.

6. The high-efficiency compression and editing method of the vehicle component load spectrum based on the Wigner-Ville transform according to claim 1, wherein The steps of using the genetic algorithm to determine the optimal threshold of the instantaneous energy spectrum include: Taking the threshold in the instantaneous energy spectrum as the design variable, taking the minimum value min(y) of the length compression ratio before and after load compression and editing as the objective function, and taking the damage ratio between the compressed signal and the original signal not less than a and the statistical parameter error less than b as the constraint conditions, establish a threshold optimization mathematical model; a and b are preset error thresholds; Use the threshold optimization mathematical model to perform threshold optimization calculation to find the optimal threshold within the numerical range [0, maxP(n)] of the instantaneous energy spectrum; Among them, the threshold optimization mathematical model is as follows: Wherein, Δavr, Δku, and Δrms are the change amounts of the mean value, kurtosis coefficient, and root mean square value respectively; L(y) is the length compression ratio before and after load compression editing; L y is the length after load compression editing, and L0 is the length before load compression editing; is the relative damage retention amount of the signal before and after editing; d(x0(t)) and d(x y (t)) are the pseudo-damage values of the signal before and after editing respectively; Among them, the pseudo-damages d(x0(t)) and d(x y (t)) of the signals before and after editing are theoretical values, without average stress correction. The damage values of each stress cycle calculated are as follows for the total pseudo-damage calculation of the load data: Where D represents the total pseudo-damage and n j is the number of cycles at the j-th stress level; N j is the number of cycles to failure at this stress level.

7. The method for efficiently compressing and editing the load spectrum of automotive parts based on the Wigner-Ville transform according to claim 1, wherein In step 7), when there is no intersection in the time series corresponding to the instantaneous energy spectrum segment in the original signal, the remaining segments in the original signal can be directly spliced to obtain the compressed signal.

8. The efficient compression and editing method of the vehicle component load spectrum based on the Wigner-Ville transform according to claim 1, characterized in that The statistical parameters include: mean value, root mean square value, and kurtosis coefficient.

9. The efficient compression and editing method of the automotive component load spectrum based on the Wigner-Ville transform according to claim 1, characterized in that The frequency-domain characteristics of the load signal are represented as a power spectral density curve, and the amplitude-domain characteristics are represented as a cross-level count distribution.

Images