A load spectrum similarity evaluation method based on overall characteristics

Abstract

This invention belongs to the field of aircraft load spectrum evaluation technology, specifically involving a load spectrum similarity evaluation method based on overall characteristics. First, two load spectra to be evaluated for similarity are selected and preprocessed. Then, multiple characteristic parameters of the load spectra are extracted from multiple dimensions. Finally, the differences in each parameter are evaluated to determine similarity. The load spectrum similarity evaluation method based on absolute metrics proposed in this invention does not rely on subjective experience but is entirely based on objective analysis. Compared with previous methods, it improves the reliability and applicability of engineering applications and provides a reliable means and basis for appropriate and reasonable service life analogy analysis.

Description

Technical Field

[0001] This invention belongs to the field of aircraft load spectrum evaluation technology, specifically relating to a load spectrum similarity evaluation method based on overall characteristics. Background Technology

[0002] Currently, the service life analogy analysis method is widely used in engineering to predict fatigue life. This method can avoid the uncertain assumption that accumulated damage reaching a certain critical value represents fatigue cracking, and it introduces existing experimental life data as the basis or reference for life prediction, thus having high life prediction accuracy.

[0003] The lifespan analogy analysis method relies on the assumptions of identical or similar structural details (same material) and similar load spectra. While identical structural details and materials are easy to understand, similar load spectra lack quantitative description. Industry professionals can only roughly classify load spectra based on their approximate uses and characteristics: generally, load spectra with similar uses and applications are considered similar. Such a classification is vague and subjective, depending on the experience of practitioners, and cannot objectively describe the degree of similarity in load spectra. In the early days when there were fewer aircraft models, this classification might have met engineering needs. However, the multi-functional and multi-purpose nature of modern aircraft models makes this previously "coarse" classification no longer suitable for current engineering applications. Therefore, a more refined and precise method is needed to replace the previous classification. Summary of the Invention

[0004] The purpose of this invention is to address the current limitations of quantitatively assessing load spectrum similarity in lifespan analogy methods. This invention provides a comprehensive and detailed analysis of load spectra from multiple dimensions and indicators, proposing various parameter criteria to satisfy load spectrum similarity, thus offering a reliable means and basis for appropriate and reasonable lifespan analogy methods.

[0005] The technical solution of the present invention: In order to achieve the above objectives, the present invention proposes a load spectrum similarity assessment method based on absolute measurement. First, two load spectra to be assessed for similarity are selected and preprocessed. Then, multiple feature parameters of the load spectra are extracted from multiple dimensions. Finally, the differences of each parameter are evaluated to determine the similarity.

[0006] Furthermore, two load spectra to be compared are selected and denoted as [X]. i ](i=1,2...n) and [Y j (j=1,2…m). Where, X i Y represents the data points in the load spectrum 1 sequence, with a sequence length of n; j denoted as data points in the load spectrum 2 sequence, with a sequence length of m.

[0007] Furthermore, for the two load spectra [X] i ] and [Y j Perform normalization preprocessing; the formula is:

[0008] [x i ] = [X i ] / NORM x (i = 1, 2, ..., n)

[0009] [y j ] = [Y j ] / NORM y (j=1,2…m)

[0010] [x i ](i=1,2...n) and [y j ](j=1,2…m) is [X i ] and [Y j The spectral sequences after normalization;

[0011] NORM x and NORM y For their respective normalized denominators: when the load types of the two load spectra are the same, then NORM x =NORM y =max{[X i ],[Y j When the load types of the two load spectra are inconsistent, NORM... x =max{[X i NORM y =max{[Y j ]}.

[0012] Furthermore, the preprocessing step involves processing the selected two load spectra [X]. i ] and [Y j Normalization preprocessing is performed; when the load types of the two load spectra are the same, the normalization denominator is the maximum value of the two load spectra; when the load types of the two load spectra are different, the normalization denominator is the maximum value of their respective load spectra.

[0013] Furthermore, multiple characteristic parameters of the load spectrum are extracted from multiple dimensions, and the amplitude skewness ΔS is extracted from a statistical dimension. skew skewness R of stress ratio skew The mean value of the stress ratio R mean ;

[0014] The average overload ratio OL of each load spectrum was extracted from the dimension of high-load hysteresis effect. mean and the distance between the unitized average maximum value and dis mean ;

[0015] The slope k1 and the mean rate of change of the slope k' of the first-order broken line segment of the two load spectra are extracted from the shape dimension of the load exceedance number curve. mean and the standard deviation of the rate of change of slope k' σ ;

[0016] From the correlation dimension of the probability pairing matrix, the probability pairing matrix of the two load spectra is operated on to extract the correlation coefficient C;

[0017] The fractal dimension D is extracted from the self-similarity dimension of the load spectrum.

[0018] Furthermore, when extracting feature parameters from the dimension of high-load hysteresis effect: average overload ratio OL mean It is the average of the ratios of several maximum values ​​in each load spectrum to its next adjacent peak value;

[0019] Unitized average maximum distance dis mean The distance from the average maximum value of each load spectrum to DIS mean The ratio to the spectral length L; where the average maximum value of the load spectrum is far from DIS. mean It is the average distance between two adjacent maximum values ​​among several maximum values ​​of the load spectrum.

[0020] Furthermore, the slope k1, mean rate of change of slope k', and standard deviation of rate of change of slope k' of the first-order piecewise linear segments of the two load spectra are extracted from the shape dimension of the load exceedance number curve. σ The extraction method is:

[0021] First, the exceedance number-load curves for each load spectrum are plotted with a load increment of 0.1 for each load level. The exceedance number-load curves are represented by broken line segments. The vertical axis represents the exceedance number, which is displayed on a logarithmic scale with a base of 10, and the horizontal axis represents the load level.

[0022] Calculate the slope of each line segment, and denote the slope of the first line segment as k1; calculate the rate of change of the slope between any two adjacent line segments, and calculate their mean and standard deviation, which are denoted as the mean rate of change of slope k'mean and the standard deviation of the rate of change of slope k', respectively. σ .

[0023] Furthermore, from the perspective of correlation of the probability pairing matrix, the probability pairing matrix of the two load spectra is operated on to extract the correlation coefficient; the method is as follows:

[0024] Construct peak and valley probability pairing matrices for each load spectrum with a load level of 0.1. Align the two probability pairing matrices to the same dimension according to the peak and valley envelopes, and calculate the correlation coefficient C between the two probability pairing matrices.

[0025] Furthermore, the fractal dimension D is extracted from the self-similarity dimension of the load spectrum, and each load spectrum sequence is regarded as a one-dimensional fractal to calculate the fractal dimension D of the respective load spectrum.

[0026] Furthermore, the extracted feature parameters are evaluated differentially;

[0027] For each characteristic parameter, corresponding evaluation index parameters are set to evaluate whether the relationship between the corresponding characteristic parameters of the two load spectra meets the set conditions. When a certain number of index parameters meet the set conditions, the two load spectra are determined to be similar; otherwise, they are not similar.

[0028] The beneficial effects of the present invention are as follows: The load spectrum similarity evaluation method based on absolute measurement proposed above does not rely on subjective experience, but is based entirely on objective analysis. Compared with the previous method, it improves the reliability and applicability of engineering applications and provides a reliable means and basis for appropriate and reasonable service life analogy analysis. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating the technical process of the present invention.

[0030] Figure 2 This is a schematic diagram showing the overload ratio and the distance to the maximum value related to parameters 4 and 5 in step four of the solution.

[0031] Figure 3 This is a schematic diagram of the exceedance number-load curves related to parameters 6, 7, and 8 in step five of the solution.

[0032] Figure 4 This is a schematic diagram of the probability pairing matrix related to parameter 9 in step six of the solution.

[0033] Figure 5 This is a schematic diagram of two overload spectrum portions (first 5000 points) selected in the embodiment;

[0034] Figure 6 This is a schematic diagram of the load spectrum amplitude distribution in the embodiment;

[0035] Figure 7 This is a schematic diagram of the load spectrum stress ratio distribution in the embodiment;

[0036] Figure 8 This is a schematic diagram of the exceedance number-load curve in the embodiment;

[0037] Figure 9 This is a partial schematic diagram of the probability pairing matrix in the embodiment;

[0038] Figure 10 This is a schematic diagram of the box-counting dimension method used in the embodiments; Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0040] A load spectrum similarity assessment method based on absolute metric is proposed. First, two load spectra to be assessed are selected and preprocessed. Then, multiple feature parameters of the load spectra are extracted from multiple dimensions. Finally, the differences between these parameters are evaluated to determine similarity. The technical process is described in [link to technical details]. Figure 1 .

[0041] The evaluation method proposed in this invention includes the following steps in its implementation:

[0042] Step 1: Select the two load spectra to be compared, denoted as [X... i ](i=1,2...n) and [Y j (j=1,2…m). Where, X i Y represents the data points in the load spectrum 1 sequence, with a sequence length of n; j For each data point in the load spectrum 2 sequence, the length of the spectrum sequence is m;

[0043] Step two, for [X] i ] and [Y j Perform normalization preprocessing; the formula is:

[0044] [x i ] = [X i ] / NORM x (i = 1, 2, ..., n)

[0045] [y j ] = [Y j ] / NORM y (j=1,2…m)

[0046] [x i ](i=1,2...n) and [y j ](j=1,2…m) is [X i ] and [Y j The spectral sequences after normalization.

[0047] NORM x and NORM y For their respective normalized denominators: when the load types of the two load spectra are the same (e.g., both are overload spectra or both are bending moment spectra), then NORMx =NORM y =max{[X i ],[Y j When the load types of two load spectra are inconsistent (e.g., one is an overload spectrum and the other is a bending moment spectrum), then NORM x =max{[X i NORM y =max{[Y j ]};

[0048] Step 3: Based on the normalization process, extract multiple characteristic parameters of the load spectrum from multiple dimensions. These characteristic parameters have typical representative meanings. The degree of similarity is objectively judged based on the differences between these typical characteristic parameter indicators.

[0049] In the specific implementation process, the skewness ΔS of the amplitude is extracted from a statistical dimension. skew skewness R of stress ratio skew The mean value of the stress ratio R mean ; these are respectively denoted as parameter 1, parameter 2, and parameter 3;

[0050] The average overload ratio OL of each load spectrum was extracted from the dimension of high-load hysteresis effect. mean and the distance between the unitized average maximum value and dis mean ; these are respectively denoted as parameter 4 and parameter 5;

[0051] The slope k1 and the mean rate of change of the slope k' of the first-order broken line segment of the two load spectra are extracted from the shape dimension of the load exceedance number curve. mean and the standard deviation of the rate of change of slope k' σ ; these are respectively denoted as parameter 6, parameter 7, and parameter 8;

[0052] From the correlation dimension of the probability pairing matrix, the probability pairing matrix of the two load spectra is operated on to extract the correlation coefficient C; denoted as parameter 9;

[0053] The fractal dimension D is extracted from the self-similarity dimension of the load spectrum; denoted as parameter 10.

[0054] In the specific implementation process, the method for extracting feature parameters 1 to 3 is as follows: [x] i ] and [y j Perform rainflow counting and calculate the amplitude ΔS of each pair of rainflow cycles for each load spectrum. i (=S max,i -S min,i ) and stress ratio R i (=S min,i / S max,i ), where S max,i and Smin,i These represent the peak and trough values ​​of the i-th rainflow cycle, respectively. The distribution of amplitude and stress ratio is statistically analyzed to obtain the amplitude skewness ΔS. skew (denoted as parameter 1) and stress ratio skewness R skew (denoted as parameter 2) and the average stress ratio R mean (Parameter 3); The formulas for calculating sample skewness and sample mean can be found in general probability and statistics books;

[0055] In the specific implementation process, the method for extracting feature parameters 4 to 5 is as follows: Calculate the average overload ratio OL of each load spectrum. mean (The mean of the ratios of the maximum value to the next peak in the spectrum, parameter 4), and the average distance of the maximum value from DIS mean The ratio of the mean distance between the maximum values ​​in the spectrum to the spectral length L is used to obtain the normalized average maximum distance dis. mean (Parameter 5), L = n or m; the distance between the overload ratio and the maximum value is as follows: Figure 2 As shown in the figure, x i and x j For [x i The two most significant values ​​in the spectrum of a closely contiguous sequence, x i Overload ratio OL ni =x i / x i+1 x j Overload ratio OL nj =x j / x j+1 x i and x j The distance is DIS nij The formula for calculating the sample mean can be found in general probability and statistics books.

[0056] In the specific implementation process, the method for extracting feature parameters 6 to 8 is as follows: Plot the exceedance number-load curve for each load spectrum with a load increment of 0.1 for each load level (only peak values ​​are counted). The coordinates of each point in the curve segment are (P... j N j ), j = 1, 2…p. Where P j P1 represents the j-th load level. <P2<…<P j <… <P p ≤1.0, N j Represents the j-th level load P j The corresponding exceedance number. Exceedance number-load curve illustration. Figure 3 Calculate the slope k of each broken line segment in the curve. j =(lgN) j -lgN j+1 ) / (P j -P j+1), j = 1, 2…p-1, and the rate of change of slope k for each segment. j '=k j+1 -k j , j=1,2…p-2; Calculate the slope k1 (parameter 6) of the first-order piecewise linear segment of the two load spectra, and the mean rate of change of slope k' mean (Parameter 7) and the standard deviation of the rate of change of slope k' σ (Parameter 8); The formulas for calculating the sample mean and sample standard deviation can be found in general probability and statistics books;

[0057] In the specific implementation process, the method for extracting feature parameter 9 is as follows: construct a peak / valley probability pairing matrix for each load spectrum with a load level of 0.1, denoted as A. n1×m1 and B n2×m2 Probability pairing matrix indicates opinions Figure 4 f in the figure i,j This represents the number of occurrences of a pairing between a valley of level i and a peak of level j. The two probability pairing matrices are aligned to the same dimension according to the peak and valley envelopes, i.e., A. n×m and B n×m Where n = max{n1, n2}, m = max{m1, m2}. Calculate the correlation coefficient C (parameter 9) between the two matrices;

[0058] The formula for calculating the correlation coefficient C is as follows:

[0059]

[0060] In the formula, tr() is the trace of the matrix, A is the probability pairing matrix of load spectrum 1, and B is the probability pairing matrix of load spectrum 2.

[0061] In the specific implementation process, the method for extracting feature parameter 10 is as follows: the load spectrum sequence is regarded as a one-dimensional fractal, and [x] is calculated respectively. i ](i=1,2...n) and [y j The fractal dimension D (parameter 10) of [j = 1, 2, ..., m]. For methods of calculating the fractal dimension, refer to general fractal geometry books.

[0062] Step four: Perform differential evaluation on the 10 feature parameters extracted above. To easily obtain intuitive evaluation results, each parameter can be evaluated in the following scoring format:

[0063] Parameter 1: When the amplitude skewness ΔS of load spectrum 1 skew,1 Amplitude skewness ΔS of load spectrum 2 skew,2 The absolute value of the difference satisfies the condition |ΔS skew,1 -ΔS skew,2 If |≤0.5, score 1 point; otherwise, score 0 points.

[0064] Parameter 2: When the stress ratio skewness R of load spectrum 1 skew,1 Stress ratio skewness R compared to load spectrum 2 skew,2 The absolute value of the difference satisfies the condition |R skew,1 -R skew,2 If |≤0.5, score 1 point; otherwise, score 0 points.

[0065] Parameter 3: When the average stress ratio R of load spectrum 1 mean,1 The average stress ratio R of load spectrum 2 mean,2 The absolute value of the difference satisfies the condition |R mean,1 -R mean,2 If |≤0.1, score 1 point; otherwise, score 0 points.

[0066] Parameter 4: When the average overload ratio of load spectrum 1 is OL mean,1 The average overload ratio OL of load spectrum 2 mean,2 Meet the conditions (OL) mean,1 -OL threshold )·(OL mean,2 -OL threshold )>0 is counted as 1 point, otherwise 0 points; where OL threshold This is the overload threshold value, which can be determined by testing. For aluminum alloys, it is generally taken as 1.3 to 1.5.

[0067] Parameter 5: When the average maximum value distance ratio of load spectrum 1 is DIS mean,1 / n and the ratio of the average maximum value distance of load spectrum 2 DIS mean,2 The ratio / m satisfies the condition 0.1≤(DIS mean,1 / n) / (DIS mean,2 1 point is awarded if ( / m)≤10, otherwise 0 points are awarded;

[0068] Parameter 6: When the initial piecewise slope k of load spectrum 1 1,谱1 The initial piecewise slope k of the load spectrum 2 1,谱2 The absolute value of the difference satisfies the condition |k 1,谱1 -k 1,谱2 If |≤0.5, score 1 point; otherwise, score 0 points.

[0069] Parameter 7: The mean rate of change of the slope of load spectrum 1, k' mean,谱1 The mean rate of change of the slope of load spectrum 2, k' mean,谱2 The absolute value of the difference satisfies the condition |k' mean,谱1 -k' mean,谱2 If |≤0.5, score 1 point; otherwise, score 0 points.

[0070] Parameter 8: The standard deviation k' of the rate of change of the slope of load spectrum 1 σ,谱1 and the standard deviation of the rate of change of the slope of load spectrum 2 σ,谱2The absolute value of the difference satisfies the condition |k' σ,谱1 -k' σ,谱2 If |≤0.5, score 1 point; otherwise, score 0 points.

[0071] Parameter 9: 1 point is awarded when the correlation coefficient C meets the condition C≥0.6, otherwise 0 points are awarded;

[0072] Parameter 10: When the absolute value of the difference between the fractal dimension D1 of load spectrum 1 and the fractal dimension D2 of load spectrum 2 satisfies the condition |D1-D2|≤0.05, 1 point is awarded; otherwise, 0 points are awarded.

[0073] In the above embodiments, it will be understood by those skilled in the art that the setting conditions for each typical characteristic parameter in the two load spectra can be other conditions. For load spectra of different aircraft models, the above setting conditions can be determined by the following method:

[0074] For parameters 1, 2, 3, 5, and 10;

[0075] These five feature parameters can be based on one of the load spectra. All peak and valley values ​​of the load spectrum are counted, and the spectrum is randomly rearranged several times. A new load spectrum is obtained after each rearrangement, which is called the test spectrum. It is recommended that the number of rearrangements be no less than 1000. For all test spectra, the corresponding parameters 1, 2, 3, 5, and 10 are extracted according to the feature extraction method of this invention. The maximum value a and minimum value b of each parameter in all test spectra are counted respectively, and (ab) / 2 is used as the setting condition for each parameter.

[0076] Among the settings corresponding to parameter 4, the overload threshold value OL threshold The value can be obtained from fatigue comparison tests; conduct two-stage constant amplitude spectrum tests with a stress ratio of 0.06. The first stage coefficient-modified peak value is S, where S≥1, and the recommended frequency is 10 times; the second stage coefficient-modified peak value is 1, and the recommended frequency is not less than 500 times; conduct group fatigue tests sequentially with S=1, S=1.1, S=1.2... When a significant high-load hysteresis effect appears, the current coefficient-modified peak value S is taken as the overload threshold value OL. threshold ;

[0077] The setting conditions corresponding to parameters 6, 7, and 8 can be based on one of the load spectra. The maximum load level of the exceedance number-load curve shifted 10% to the left along the x-axis is taken as the left boundary; the maximum load level shifted 10% to the right along the x-axis is taken as the right boundary. For each exceedance number, the corresponding load level is randomly selected within the left and right boundaries to obtain a new test exceedance number curve, and this process is repeated several times, with a suggested number of repetitions not less than 1000. For all test exceedance number curves, the corresponding parameters 6, 7, and 8 are extracted according to the method proposed in this invention. The maximum value c and minimum value d of each parameter in all test exceedance number curves are calculated respectively, and (cd) / 2 is used as the setting condition for each parameter.

[0078] Regarding parameter 9, it is generally believed in the industry that when C≥0.6, the two matrices are correlated; therefore, 0.6 is used as the setting condition.

[0079] Step 5: Based on the differential evaluation score, make a similarity judgment.

[0080] The maximum score is 10 points: 8-10 points, highly similar; 6-7 points, similar; 3-5 points, not very similar; 0-2 points, almost no similarity.

[0081] When the score is ≥6, the service life analogy analysis method is reliable and appropriate.

[0082] It should be noted that the above scoring is only a simplified form of evaluation. Obviously, for those skilled in the art, scoring is only one intuitive form of representation and is not limited to it. There are several other implementation forms, such as using color depth for evaluation. Finally, the color superposition depth can also intuitively reflect the similarity between two load spectra. There can be many more similar specific evaluation measures, which will not be elaborated on in this invention.

[0083] Example 1: Taking two overload spectra as an example, the load spectrum similarity evaluation method based on absolute metric proposed in this invention is adopted. The technical process is as follows: Figure 1 One possible specific implementation includes the following steps:

[0084] The first step is to select two load spectra for similarity measurement. In this embodiment, two overload spectra are selected as the evaluation objects, denoted as [Nz], ... 1i ](i=1,2...n) and [Nz 2j (j=1,2…m); the first 5000 points of the spectral sequence indicate the following. Figure 5 ;

[0085] The second step is to normalize the two load spectra from the first step. Since both load spectra represent overloads, the denominator for the normalization of the two load spectra is Nz. max=max{[Nz i ],[Nz j The normalized spectral sequence is denoted as [nz]. 1i ](i=1,2...n) and [nz 2j (j = 1, 2…m);

[0086] The third step is to extract feature parameters 1 to 3 from the two normalized sequences in the second step:

[0087] Parameter 1 distribution see Figure 6 Statistical analysis yields ΔS skew,1 =1.238, ΔS skew,2 =0.656;

[0088] See parameter 2 distribution. Figure 7 Statistical analysis yields R skew,1 = -1.307, R skew,2 =0.446;

[0089] See parameter 3 distribution. Figure 7 Statistical analysis yields R mean,1 =0.20, R mean,2 =0.42;

[0090] Fourth step: Extract feature parameters 4 to 5 from the two normalized sequences in the second step.

[0091] Parameter 4, statistically, yields OL mean,1 =1.549, OL mean,2 =1.638;

[0092] Parameter 5, statistically, yields DIS mean,1 / n=0.0165, DIS mean,2 / m=0.1042;

[0093] Fifth step: Extract feature parameters 6 to 8 from the two normalized sequences in the second step.

[0094] Parameter 6 indicates opinion Figure 8 Statistical analysis yields k 1,谱1 =-1.43, k 1,谱2 = -1.86;

[0095] Parameter 7 indicates opinion Figure 8 Statistical analysis yields k' mean,谱1 =-0.70, k' mean,谱2 = -0.67;

[0096] Parameter 8 indicates opinion Figure 8 Statistical analysis yields k' σ,谱1 =0.53, k' σ,谱2=0.85;

[0097] Step 6: Extract feature parameter 9 from the two normalized sequences in step 2.

[0098] Parameter 9 indicates opinion Figure 9 The calculation yields C = 0.72;

[0099] Step 7: Extract feature parameter 10 from the two normalized sequences in step 2.

[0100] Parameter 10 is calculated and obtained using the box-counting dimension method in this embodiment. The method is described in the following figure. Figure 10 The calculation formula is:

[0101]

[0102] The calculation yields D1 = 1.87 and D2 = 1.89. Other methods can also be used to obtain these values.

[0103] Step 8: Based on the parameter statistics results from step 3, conduct a differential evaluation, as detailed in Table 1.

[0104] Table 1

[0105]

[0106] Step 9: According to Table 1, the similarity evaluation score of spectrum 1 and spectrum 2 is 7 points, which means that they are basically similar. Lifetime analogy analysis can be carried out based on spectrum 1 and spectrum 2.

[0107] This invention starts from the overall characteristics of the load spectrum, proposes various characteristic parameters from multiple dimensions to describe the characteristics of the load spectrum, and judges the degree of similarity based on the differences between the parameter indicators. It establishes a quantitative evaluation method for the similarity between load spectra. Based on this method, a quantitative judgment of the degree of similarity between load spectra can be effectively given.

[0108] The above description is merely a specific embodiment of the present invention, providing a detailed description of the invention. Parts not covered herein are conventional techniques. However, the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for load spectrum similarity assessment based on absolute metrics, characterized in that, The evaluation method first selects two load spectra to be evaluated for similarity and preprocesses them; then extracts multiple feature parameters of the load spectra from multiple dimensions; finally, it evaluates the differences of each parameter and makes a similarity judgment. Multiple feature parameters of the load spectrum are extracted from multiple dimensions, including the skewness Δ of the amplitude extracted from a statistical dimension. S skew skewness of stress ratio R skew The mean of the stress ratio R mean The average overload ratio of each load spectrum was extracted from the dimension of high-load hysteresis effect. OL mean and the distance between the unitized average and the maximum value dis mean The slopes of the first-order piecewise linear segments of the two load spectra were extracted from the shape dimension of the load exceedance number curve. k 1. Mean rate of change of slope and the standard deviation of the rate of change of slope From the perspective of correlation of the probability pairing matrix, the probability pairing matrix of the two load spectra is operated on to extract the correlation coefficient. C The fractal dimension is extracted from the self-similarity dimension of the load spectrum. D When extracting feature parameters from the dimension of high-load hysteresis effect: average overload ratio OL mean The mean of the ratios of several maximum values ​​in each load spectrum to the next adjacent peak; the unitized average distance to the maximum value. dis mean Distance between the average maximum value of each load spectrum DIS mean With spectral length L The ratio; where the average maximum value of the load spectrum is far from DIS mean It is the average distance between two adjacent maximum values ​​among several maximum values ​​of the load spectrum; The slopes of the first-order broken line segments of the two load spectra were extracted from the shape dimension of the load exceedance number curve. k 1. Mean rate of change of slope and the standard deviation of the rate of change of slope The extraction method is: First, the exceedance number-load curves for each load spectrum are plotted with a load increment of 0.1 for each load level. The exceedance number-load curves are represented by broken line segments. The vertical axis represents the exceedance number, which is displayed on a logarithmic scale with a base of 10, and the horizontal axis represents the load level. Calculate the slope of each line segment, and denote the slope of the first line segment as... k 1. Calculate the rate of change of the slope of any two adjacent broken line segments, and calculate their mean and standard deviation, which are denoted as the mean rate of change of slope. and the standard deviation of the rate of change of slope ; From the correlation dimension of the probability pairing matrix, the correlation coefficient is extracted by operating on the probability pairing matrices of the two load spectra; the method is as follows: Construct peak and trough probability pairing matrices for each load spectrum with a load level of 0.

1. Align the two probability pairing matrices to the same dimension according to the peak and trough envelopes, and calculate the correlation coefficient between the two probability pairing matrices. C ; The fractal dimension is extracted from the self-similarity dimension of the load spectrum. D， Treating each load spectrum sequence as a one-dimensional fractal, calculate the fractal dimension of each load spectrum. D ; The extracted feature parameters are evaluated differentially. For each feature parameter, corresponding evaluation index parameters are set to evaluate whether the relationship between the corresponding feature parameters of the two load spectra meets the set conditions. When a certain number of index parameters meet the set conditions, the two load spectra are judged to be similar; otherwise, they are not similar.

2. The load spectrum similarity assessment method based on absolute metric as described in claim 1, characterized in that, Select two load spectra to be compared, denoted as [ X i ]and[ Y j ],in, i =1,2…n, j =1,2…m, X i For each data point in the load spectrum 1 sequence, the length of the spectrum sequence is n; Y j denoted as data points in the load spectrum 2 sequence, with a sequence length of m.

3. A load spectrum similarity assessment method based on an absolute metric as defined in claim 2, characterized in that, Two load spectra [ X i ] and [ Y j ] were normalized and pre-processed; the formula is [ x i ]=[ X i ] / NORM x , i =1,2…n; [ y j ]=[ Y j ] / NORM y , j =1,2…m; [ x i ]and[ y j ]for X i and Y j Normalized spectral sequences; NORM x and NORM y For their respective normalized denominators: when the load types of the two load spectra are the same, then NORM x =NORM y =max{[ X i ],[ Y j When the load types of the two load spectra are inconsistent, NORM... x =max{[ X i NORM y =max{[ Y j ]}.

Images