Rainflow counting fatigue analysis method based on sliding window algorithm

Abstract

The application provides a rain flow counting fatigue analysis method based on a sliding window algorithm. The method comprises obtaining time series load data and selecting subsequence data according to a preset sliding window value. Then, the hysteresis loop closure of the subsequence data is judged by the sliding window algorithm. If the subsequence data meets the hysteresis loop closure rule, the subsequence data is extracted as a rain flow pair, the sliding window is moved according to a first sliding window strategy, and the next round of hysteresis loop closure judgment is performed. If the subsequence data does not meet the hysteresis loop closure rule, the sliding window is moved according to a second sliding window strategy, and the next round of hysteresis loop closure judgment is performed. In this way, the complexity of rain flow counting can be reduced, the extraction efficiency of rain flow pairs can be improved, and the actual needs of complex engineering applications can be met.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of engineering machinery, and in particular to a rainflow counting fatigue analysis method based on a sliding window algorithm. BACKGROUND

[0002] Structural fatigue failure is a serious challenge in the engineering field. In key fields such as aerospace, automotive engineering, wind power, engineering machinery, and structural health monitoring, it is crucial to accurately assess the fatigue life of components under complex alternating loads. Rainflow counting (RFC) is a core means of extracting equivalent cycles of fatigue damage, and has become an indispensable tool for modern fatigue life prediction. However, in the face of ultra-long time series data and real-time requirements generated by modern engineering, traditional RFC methods still have serious limitations. First, the loading of ultra-long load sequences increases the memory requirements of rainflow counting, far exceeding the resource limits of embedded systems or edge devices. Second, the traditional RFC method has high computational complexity, and in continuous monitoring scenarios that require frequent updates of analysis results, computational delay becomes a performance bottleneck, hindering the possibility of dynamic response. In addition, the batch processing mode in the traditional RFC method cannot adapt to continuous input data streams, and each update requires re-computation of the entire sequence, resulting in extremely low computational efficiency and serious resource waste.

[0003] Therefore, there is an urgent need for a rainflow counting fatigue analysis method that can efficiently process ultra-long sequences, significantly reduce memory dependence, and support real-time analysis. SUMMARY

[0004] According to embodiments of the present application, a rainflow counting fatigue analysis scheme based on a sliding window algorithm is provided, which can effectively reduce the complexity of rainflow counting and significantly improve the extraction efficiency of rainflow pairs.

[0005] In a first aspect of the present application, a rainflow counting fatigue analysis method based on a sliding window algorithm is provided. The method comprises:

[0006] Obtaining time series load data, and selecting sub-sequence data according to a preset sliding window value;

[0007] Performing hysteresis loop closure judgment on the sub-sequence data by a sliding window algorithm;

[0008] If the sub-sequence data meets the hysteresis loop closure rule, the sub-sequence data is extracted as a rainflow pair, the sliding window is moved according to a first sliding window strategy, and the next round of hysteresis loop closure judgment is performed;

[0009] If the sub-sequence data does not meet the hysteresis loop closure rule, the sliding window is moved according to a second sliding window strategy, and the next round of hysteresis loop closure judgment is performed.

[0010] In a possible implementation, before the time-series load data is acquired, the method further includes,

[0011] acquiring multi-condition load sequence data, and performing extreme point monitoring and screening on the multi-condition load sequence data;

[0012] removing non-peak-valley points in the multi-condition load sequence data to acquire the time-series load data.

[0013] In a possible implementation, the hysteresis loop closure rule includes,

[0014] the amplitude between adjacent peak-valley points of the load data in the subsequence data is less than the amplitude between adjacent peak-valley points of the load data in the two adjacent subsequence data.

[0015] In a possible implementation, the first sliding window strategy includes,

[0016] moving the sliding window backward according to a preset sliding window value to acquire next subsequence data;

[0017] the next subsequence data sequentially performs data coverage on the subsequence data.

[0018] In a possible implementation, the second sliding window strategy includes,

[0019] moving the sliding window by one unit backward, and performing data coverage on the subsequence data by the in-situ updating strategy.

[0020] In a possible implementation, the method further includes,

[0021] after one round of traversal of the sliding window algorithm is completed, detecting and verifying residual values in the time-series load data that are not included in the sliding window algorithm for calculation, and determining whether there is end data missing.

[0022] Optionally, the detecting and verifying the residual values in the time-series load data that are not included in the sliding window algorithm for calculation includes,

[0023] determining whether the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is equal to the length of the time-series load data;

[0024] if the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is equal to the length of the time-series load data, ending the traversal of the sliding window algorithm;

[0025] if the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is not equal to the length of the time-series load data, supplementing the residual values to the end of the next round of traversal of the sliding window algorithm.

[0026] In a possible implementation, the method further includes,

[0027] After extracting the sub-sequence data into the rainflow pair, the rainflow pair is merged into a global list, and the fatigue damage information is updated in real time according to the extracted rainflow pair.

[0028] In a possible implementation, the method further includes,

[0029] When the sliding window algorithm completes the entire history traversal, the obtained global fatigue damage information is closed-loop verified according to preset fatigue damage reference data, and the rainflow counting reliability is verified.

[0030] In a possible implementation, the method further includes,

[0031] The sub-sequence data is dynamically managed by a memory in-place update mechanism.

[0032] The rainflow counting fatigue analysis method based on the sliding window algorithm provided by the embodiments of the present application acquires time sequence load data, then selects sub-sequence data according to a preset sliding window value, and then performs hysteresis loop closure judgment on the sub-sequence data by using the sliding window algorithm. If the sub-sequence data meets the hysteresis loop closure rule, the sub-sequence data is extracted into a rainflow pair, and the sliding window is moved according to the first sliding window strategy, and the next round of hysteresis loop closure judgment is performed. If the sub-sequence data does not meet the hysteresis loop closure rule, the sliding window is moved according to the second sliding window strategy, and the next round of hysteresis loop closure judgment is performed. Efficient conversion from original time domain signals to fatigue damage evaluation is realized, and frequent data operations are reduced.

[0033] It should be understood that the content described in the summary section is not intended to limit the key or important features of the embodiments of the present application, nor to limit the scope of the present application. Other features of the present application will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0034] The above and other features, advantages, and aspects of the embodiments of the present application will become more apparent by describing in detail the following embodiments with reference to the attached drawings. In the drawings, the same or similar reference numerals refer to the same or similar elements, and:

[0035] Figure 1 A flowchart of a rainflow counting fatigue analysis method based on a sliding window algorithm according to an embodiment of the present application;

[0036] Figure 2 A schematic diagram of a random load stress spectrum according to an embodiment of the present application;

[0037] Figure 3 A schematic diagram of a first sliding window strategy according to an embodiment of the present application;

[0038] Figure 4 a schematic diagram of a second sliding window strategy according to an embodiment of the present application;

[0039] Figure 5 a flowchart of a rainflow counting fatigue analysis method based on a sliding window algorithm according to an embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the objects, technical solutions and advantages of the embodiments of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are some but not all of the embodiments of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present disclosure.

[0041] In addition, the term "and / or" herein is only to describe the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " herein generally represents an "or" relationship between the front and rear associated objects.

[0042] Figure 1 A flowchart of a rainflow counting fatigue analysis method based on a sliding window algorithm according to an embodiment of the present disclosure is shown. Referring to Figure 1 , the method comprises:

[0043] S101, obtaining time sequence load data, and selecting subsequence data according to a preset sliding window value.

[0044] In the embodiment, the time sequence load data in each subsequence data is ensured to be valid load data by the sliding window, which provides a basis for subsequent rainflow calculation.

[0045] Optionally, before the time sequence load data is obtained, the method further comprises:

[0046] obtaining multi-condition load sequence data, and performing extreme point monitoring and screening on the multi-condition load sequence data;

[0047] removing non-peak-valley points in the multi-condition load sequence data to obtain the time sequence load data.

[0048] Figure 2 A schematic diagram of a random load stress spectrum according to an embodiment of the present application is shown as Figure 2 . Wherein, the horizontal axis is time sequence (second), and the vertical axis is load stress amplitude (MPa) of multi-condition load. From Figure 2As can be seen, the multi-condition load sequence data presents a random distribution characteristic. In rainflow counting, only the maximum values (peaks) and minimum values (valleys) of the multi-condition load sequence data can represent the key turning points of the load borne by the structure. These turning points determine the change direction and amplitude of the structural stress and strain. Therefore, it is also necessary to monitor and screen the extreme points of the multi-condition load data. In this application, a local extreme detection algorithm is used to identify the maximum values (peaks) and minimum values (valleys) in the multi-condition load data. For example, there is a multi-condition load sequence data x = [xi-2, xi-1, xi, xi+1, xi+2]. If xi is greater than the values of all other data points in the sequence x, that is, xi > xi-2, xi > xi-1, xi > xi, xi > xi+1, xi > xi+2, then xi is determined to be a maximum value (peak) point. If xi is less than the values of all other data points in the sequence x, that is, xi < xi-2, xi < xi-1, xi < xi, xi < xi+1, xi < xi+2, then xi is determined to be a minimum value (valley) point. Finally, after traversing the entire multi-condition load data, the non-peak-valley value points in the multi-condition load sequence data are removed, and the preprocessed time-series load data can be obtained, thereby reducing the unnecessary calculation amount in the subsequent rainflow counting and improving the rainflow counting efficiency.

[0049] In addition, the amplitude between adjacent peaks and valleys between the maximum values (peaks) and minimum values (valleys) can also be calculated to provide key data for subsequent rainflow counting. The specific calculation formula is as follows:

[0050] ,

[0051] where, [[ID=ll]] is the amplitude between adjacent peak-valley value pairs between, is the i-th maximum value (peak), is the i-th minimum value (valley), i < N, and N is the number of time-series loads

[0052] In this embodiment, by preprocessing the multi-condition load sequence data, the data processing process in the subsequent rainflow counting is simplified. Moreover, by setting the number of peak-valley value points in the time-series load data to an odd number and setting the head and tail as maximum values (peaks) and minimum values (valleys) at the same time, a basis for subsequent rainflow counting can be provided.

[0053] S102, use the sliding window algorithm to judge the hysteresis loop closure of the subsequence data.

[0054] In this embodiment, in combination with the dynamic moving characteristics of the sliding window, the rapid switching calculation and analysis are performed between different sub-sequence data segments, the hysteresis loop information existing in the data can be captured in time, and the efficiency and accuracy of the hysteresis loop closing judgment are effectively improved.

[0055] Optionally, the hysteresis loop closing rule comprises,

[0056] The amplitude between the adjacent peak and valley of the load data in the sub-sequence data is less than the amplitudes between the adjacent peak and valley of the front and rear load data.

[0057] Specifically, the determination formula of the hysteresis loop closing rule is as follows:

[0058]

[0059] Among them, is the amplitude between the adjacent peak and valley of the continuous load data. For example, the sliding window algorithm loads three continuous sliding windows, the pointer points to the i th load data, and the amplitude between the adjacent peak and valley of the three continuous load data starting from the i th load data can be obtained by calculation. If the hysteresis loop closing rule is met, the i th and i+1 th load data can constitute a complete rainflow pair.

[0060] In this embodiment, the rainflow pair in the time sequence load data is identified by the hysteresis loop closing judgment, which lays a foundation for subsequent counting.

[0061] S103, if the sub-sequence data meets the hysteresis loop closing rule, the sub-sequence data is extracted as a rainflow pair, the sliding window is moved according to the first sliding window strategy, and the next round of hysteresis loop closing judgment is performed.

[0062] In this embodiment, when the sub-sequence data meets the hysteresis loop closing rule, the rainflow pair is extracted in time, which provides a basis for subsequent fatigue analysis. In addition, by moving the sliding window through the first sliding window strategy, the continuity of data processing is ensured while avoiding repeated and invalid judgment areas.

[0063] Optionally, the first sliding window strategy comprises,

[0064] Moving the sliding window backward according to a preset sliding window value to obtain the next sub-sequence data;

[0065] The next sub-sequence data sequentially covers the sub-sequence data.

[0066] The preset sliding window value can be set as the number of load pairs identified before the current load data, or can be customized.

[0067] Figure 3 ​A schematic diagram of the first sliding window strategy according to the embodiment of the present application is shown in FIG. 1. Figure 3

[0068] The 15 load data in the subsequence data are subjected to hysteresis loop closure judgment, and the load data B (time sequence 11) to the load data C (time sequence 14) meet the hysteresis loop closure rule, that is, constitute a complete fatigue cycle. After the rain flow pairs of the load data B to C are extracted by the sliding window algorithm and the corresponding amplitude data are recorded, the sliding window needs to be moved backward for the next hysteresis loop closure judgment. In order to avoid frequent memory allocation and release, the data coverage is adopted in the present application, that is, the load data C is covered to the load data A (time sequence 10) before the next subsequence data is subjected to the hysteresis loop closure judgment, and the next subsequence data after the load data C is subjected to corresponding data coverage in turn, so as to complete the movement of the sliding window.

[0069] In the present embodiment, the first sliding window strategy is adopted, which eliminates the overhead of dynamic memory management and reduces the memory operation time.

[0070] In the present embodiment, the first sliding window strategy is adopted, which eliminates the overhead of dynamic memory management and reduces the memory operation time.

[0071] In the present embodiment, when the subsequence data does not meet the hysteresis loop closure rule, the sliding window is moved according to the second sliding window strategy, and the data region that does not meet the requirement is quickly skipped, so as to avoid wasting the calculation resources in the invalid region.

[0072] Optionally, the second sliding window strategy includes,

[0073] The sliding window is moved backward by one unit, and the subsequence data is subjected to data coverage by the in-situ updating strategy.

[0074] Figure 4 A schematic diagram of the second sliding window strategy according to the embodiment of the present application is shown in FIG. 2. Figure 4

[0075] The load data DEF in the sliding window does not meet the hysteresis loop closure rule, and the next subsequence data is obtained by moving backward by one unit. The subsequence data is subjected to data coverage by the in-situ updating strategy, that is, the load data EFG is moved to the load data DEF in turn for corresponding data coverage, so as to complete the movement of the sliding window.

[0076] In the present embodiment, the second sliding window strategy is adopted, which guarantees the continuity of the load data and avoids the memory rearrangement overhead caused by the traditional deletion operation. ​​

[0077] In a possible implementation, the method further includes,

[0078] When the sliding window algorithm completes a round of traversal, the residual values in the time series load data that are not included in the sliding window algorithm are detected and verified to determine whether there is missing end data.

[0079] In this embodiment, by detecting and verifying the residual values in the time series load data that are not included in the sliding window algorithm, the possible missing key data is found in time, and a basis is provided for subsequent data verification.

[0080] Optionally, the detecting and verifying the residual values in the time series load data that are not included in the sliding window algorithm includes,

[0081] determining whether the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is equal to the length of the time series load data;

[0082] If the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is equal to the length of the time series load data, the traversal of the sliding window algorithm is ended;

[0083] If the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is not equal to the length of the time series load data, the residual values are supplemented to the end of the next round of traversal of the sliding window algorithm.

[0084] For example, the time series load data includes 50 load data points, the sliding window size is set to 8, and after one round of traversal of the sliding window algorithm, 16 rainflow pairs are successfully extracted, that is, 48 load data points are extracted as rainflow pairs. At this time, there are 3 residual values that are not included in the calculation, and the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is not equal to the length of the time series load data. Therefore, the 3 residual values that are not included in the calculation need to be supplemented to the end of the next round of traversal of the sliding window algorithm to realize complete traversal of the load data. Compared with the scheme of deleting the load data points that have been counted one by one after identifying the loop in the traditional rainflow counting method, the present application uses the data covering method to fill the untraversed load data points to the subsequent position of the time series load history, covers the invalid data, and shortens the data length of each traversal.

[0085] In this embodiment, by using the rigorous detection and verification mechanism, the completeness and accuracy of the load data analysis are ensured, and the situation that the rainflow counting fatigue analysis result is deviated due to the unprocessed part of the data is avoided.

[0086] Optionally, the method further includes,

[0087] After extracting the sub-sequence data into rainflow pairs, the rainflow pairs are merged into a global list, and the fatigue damage information is updated in real time according to the extracted rainflow pairs.

[0088] In this embodiment, the sliding window algorithm is combined, and after the sub-sequence data is extracted into rainflow pairs, the rainflow pairs are merged into a global list and the fatigue damage information is updated in real time, which can efficiently integrate load data and dynamically reflect fatigue state, thereby improving the efficiency of fatigue analysis.

[0089] In a possible implementation, the method further includes,

[0090] After the sliding window algorithm completes the entire process traversal, the obtained global fatigue damage information is closed-loop verified according to preset fatigue damage reference data, and the rainflow counting reliability is verified.

[0091] The following Table 1 shows a load data error example table, where element is the load data number, Damage is the cumulative fatigue damage value of the load data point after rainflow counting, Life is the fatigue life predicted according to the cycle number of the load data point and the Damage value, nCode is the fatigue damage value (i.e., fatigue damage reference data) calculated by a fatigue analysis software (such as nCode DesignLife, etc.), and error is the error value between Life and nCode.

[0092] Table 1

[0093] element load data number Damage accumulated fatigue damage value Life predicted fatigue life nCode calculated fatigue damage value error (%) error value 4719 0.957432 1.04446 0.9573 -0. 013788781 4666 0. 0601548 16.6238 0.06015 -0.00798005 4785 0. 0127267 78.5752 0. 01273 0. 025923016 4621 4.72E-05 21186. 7 0.0000472 0.001059322 4853 2. 91E-05 34382 0.00002909 0. 017188037 15105 7. 41E-07 1.35E+06 7. 406E-07 -0.000675128 4668 3. 00E-08 3.33E+07 3. 001E-08 -0.004665112 15108 2.39E-08 4.18E+07 2.394E-08 -0.010442774 4720 2.39E-08 4.18E+07 2.392E-08 -0.012541806 4733 7. 16E-09 1.40E+08 7. 155E-09 -0.001537386 15115 6.94E-09 1.44E+08 6.944E-09 0. 004176267 4815 4.47E-09 2.24E+08 4.466E-09 -0.009404389 4623 3.49E-09 2.87E+08 3.489E-09 -0.006305532 4786 2.03E-09 4.92E+08 2.03E-09 -0. 024137931 ... ... ... ... ...

[0094] In this embodiment, by comparison and verification, the error value between the predicted fatigue life and the preset fatigue damage reference data is obtained, and the reliability of rainflow counting is ensured.

[0095] Optionally, the method further includes,

[0096] The sub-sequence data is dynamically managed by an in-memory update mechanism.

[0097] First, the application adopts a hybrid layout combining columnar storage and cache line accurate alignment, and stores the amplitude, mean value and other key parameters of the load data in independent continuous memory areas. Moreover, the starting address of each sub-sequence data array is set according to an integer multiple of the processor cache line size (such as 64 bytes under x86 architecture), thereby improving the cache hit rate and avoiding performance loss caused by cross-cache line access.

[0098] Secondly, the memory in-place update mechanism allocates a fixed-size ring buffer in advance by analyzing the characteristics of the load data, and uses the pointer sliding window counting to realize data in-place update. When the hysteresis loop is closed, instead of deleting the data as in the traditional method, the write pointer is moved to make the new data directly cover the processed area, which eliminates the overhead of dynamic memory management and reduces the memory operation time.

[0099] Finally, the SIMD (Single Instruction Multiple Data) instruction set can also be optimized to ensure 128 / 256-bit alignment of the key data structure, so that the AVX / AVX2 (Advanced Vector Extensions) in the CPU can fully exert its performance. At the same time, the load data is partitioned, and the load data sequence is divided into multiple blocks that can be processed independently. Each block is equipped with an independent buffer and a state machine, so that the sliding window algorithm can be transplanted to a multi-core processor or GPU architecture to achieve parallel acceleration.

[0100] In this embodiment, the cache hit rate is improved by columnar storage and accurate alignment with the cache line, and the overhead of dynamic memory management is eliminated by using the "memory in-place update" mechanism in the hysteresis loop processing algorithm, thereby improving the system performance.

[0101] Figure 5 The flowchart of the rainflow counting fatigue analysis method based on the sliding window algorithm according to the embodiments of the present application is shown in Figure 5

[0102] ​Firstly, the time sequence load data is acquired, a subsequence data is selected according to a pre-set sliding window, hysteresis loop judgment is performed on the subsequence data by combining the sliding window algorithm, if the current subsequence data meets the hysteresis loop closure rule, the current subsequence data is extracted as a rainflow pair, and the load data meeting the hysteresis loop closure rule, i.e., the amplitude and other data, is saved to a global list of fatigue damage, then the sliding window is moved according to the first sliding window strategy, and the next round of hysteresis loop closure judgment is performed, if the current subsequence data does not meet the hysteresis loop closure rule, the sliding window is moved according to the second sliding window strategy, and the next round of hysteresis loop closure judgment is performed. Then, after completing one round of iteration, it is judged whether the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is equal to the length of the time sequence load data, if the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is equal to the length of the time sequence load data, the iteration of the sliding window algorithm is ended, if the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is not equal to the length of the time sequence load data, the residual values are supplemented to the end of the next round of iteration of the sliding window algorithm, and the next round of iteration is continued, until the sum of the number of residual values and the number of rainflow pairs extracted by the sliding window algorithm is equal to the length of the time sequence load data, the iteration of the sliding window algorithm is ended, and the rainflow counting fatigue analysis of the entire time sequence load data is completed.

[0103] According to the embodiments of the present disclosure, the following technical effects are achieved:

[0104] 1) The sliding window algorithm is combined with the hysteresis loop closure judgment in the rainflow counting method, the calculation complexity is reduced, and the extraction efficiency of the rainflow pair is improved;

[0105] 2) The data coverage is used to replace the traditional dynamic memory allocation, the memory operation time is reduced, and the processing speed of the rainflow calculation is improved;

[0106] 3) The integrity check of the load sequence data is realized through the residual closed loop check, the end missing problem of the traditional algorithm is solved, and the reliability of the fatigue analysis result in the engineering application is ensured.

[0107] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the action order described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0108] The above is the introduction of the method embodiments, and the scheme described in the present application will be further described through the device embodiments below.

[0109] The computer program product of the present application can be a computer program embodied on a non-transitory computer readable medium. Such non-transitory computer readable medium can include, but is not limited to, floppy diskettes, CD-ROMs, DVDs, flash memories, memory sticks, and hard drives.

[0110] The units or modules described in the embodiments of the present application can be implemented by software or hardware. The described units or modules can also be implemented in a processor. In some cases, the names of the units or modules do not constitute a limitation on the units or modules themselves.

[0111] The above description is merely illustrative of the principles of the application. Depending upon the implementation, other techniques of accomplishing system embodiment and / or method embodiments can be employed without departing from the scope of the application. For example, the above described example has utilized a modular software architecture. Alternatively, the modules can be provided as one or more software applications or program modules. Further, the present application can be implemented as a program product such as a downloadable software application.

Claims

1. A rainflow counting fatigue analysis method based on a sliding window algorithm, characterized in that, The method comprises: obtaining time sequence load data, and selecting subsequence data according to a preset sliding window value; performing hysteresis loop closure judgment on the subsequence data by using a sliding window algorithm; if the subsequence data meets a hysteresis loop closure rule, extracting the subsequence data as a rainflow pair, moving the sliding window according to a first sliding window strategy, and performing next round hysteresis loop closure judgment; if the subsequence data does not meet the hysteresis loop closure rule, moving the sliding window according to a second sliding window strategy, and performing the next round hysteresis loop closure judgment; the hysteresis loop closure rule comprises: amplitudes between adjacent peaks and valleys of load data in the subsequence data are simultaneously smaller than amplitudes between adjacent peaks and valleys of two load data before and after the subsequence data; wherein, a judgment formula of the hysteresis loop closure rule is as follows: ， wherein, is the amplitude between adjacent peaks and valleys of the continuous load data; the method further comprises: after one round of the sliding window algorithm is completed, detecting and verifying residual values in the time sequence load data which are not included in the sliding window algorithm, and judging whether there is end data missing; the detecting and verifying the residual values in the time sequence load data which are not included in the sliding window algorithm comprises: judging whether a sum of the residual value quantity and a rainflow pair quantity extracted by the sliding window algorithm is equal to a time sequence load data length; if the sum of the residual value quantity and the rainflow pair quantity extracted by the sliding window algorithm is equal to the time sequence load data length, ending the round of the sliding window algorithm; if the sum of the residual value quantity and the rainflow pair quantity extracted by the sliding window algorithm is not equal to the time sequence load data length, supplementing the residual values to an end of next round of the sliding window algorithm; the method further comprises: after the subsequence data is extracted as the rainflow pair, merging the rainflow pair to a global list, and updating fatigue damage information in real time according to the extracted rainflow pair; the method further comprises: after the sliding window algorithm completes all rounds of the sliding window algorithm, performing closed loop verification on obtained global fatigue damage information according to preset fatigue damage reference data, and verifying rainflow counting reliability; the method further comprises: performing dynamic memory management on the subsequence data by using an in-memory update mechanism comprises: adopting a hybrid layout combining column storage and cache line accurate alignment, and storing amplitudes and mean values of load data in independent continuous memory areas respectively; and, starting addresses of arrays of each subsequence data are set according to integer multiples of processor cache line sizes; the in-memory update mechanism pre-allocates a fixed-size ring buffer by pre-analyzing characteristics of load data, and implements data in-place update by using a pointer sliding window counting, and moves a write pointer when detecting hysteresis loop closure, so that new data directly covers the processed area; optimizing for a single instruction multiple data stream instruction set; performing data partitioning on load data, and segmenting load data sequences into multiple independently processable blocks, and each block is equipped with an independent buffer and a state machine.

2. The rainflow counting fatigue analysis method based on a sliding window algorithm according to claim 1, characterized in that, before the time sequence load data is obtained, the method further comprises: Obtaining multi-working condition load sequence data, performing extreme point monitoring and screening on the multi-working condition load sequence data; Removing non-peak-valley points in the multi-working condition load sequence data to obtain the time sequence load data.

3. The rainflow counting fatigue analysis method based on a sliding window algorithm according to claim 1, characterized in that, The first sliding window strategy comprises, Moving a sliding window backward according to the preset sliding window value to obtain next sub-sequence data; The next sub-sequence data sequentially performs data covering on the sub-sequence data.

4. The rainflow counting fatigue analysis method based on a sliding window algorithm according to claim 1, characterized in that, The second sliding window strategy comprises, The sliding window moves one unit backward, and the sub-sequence data is covered by the in-situ updating strategy.

Images