Data analysis method and system for an aviation accessory measurement device system

Abstract

This application relates to the field of measurement data analysis, and particularly to a data analysis method and system for an aviation accessory measurement device system. The method includes: obtaining compressor surface vibration data within an initial time window, and obtaining a time-frequency diagram through a frequency domain analysis algorithm; calculating a surge time window adaptation coefficient corresponding to the current time according to the time-frequency diagram, and adjusting the length of the initial time window according to the surge time window adaptation coefficient to obtain the optimal length of the next moment window. This application has the effect of significantly reducing the detection delay of abnormal signals.

Description

Technical Field

[0001] This application relates to the field of measurement data analysis, and particularly to a data analysis method and system for an aviation accessory measurement device system. Background Art

[0002] Aviation accessories refer to various auxiliary devices installed on an aircraft or an aero-engine to ensure its normal operation, and usually include: engine accessories such as generators, starters, fuel pumps, lubricating oil pumps, hydraulic pumps, ice protection systems, etc.; aircraft accessories such as control systems, environmental control systems, hydraulic systems, fuel systems, electrical systems, etc.; other accessories such as wheels, propellers, landing gears, etc.

[0003] Surge is a serious aerodynamic instability phenomenon in an aero-engine, manifested as low-frequency and high-amplitude oscillations of the air flow along the axis of the compressor. The main reasons for its occurrence are: the actual operating flow rate is less than the surge flow rate, such as excessive production reduction, insufficient intake air source, blocked inlet filter, etc.; or the outlet pressure of the compressor is lower than the pipeline network pressure, resulting in the operating working point moving towards the small flow rate area. In addition, external disturbances such as component aging, thunderstorm weather, sand and dust may also trigger surge. Surge is extremely harmful to an aero-engine, which can cause the intake duct or tail nozzle to fire, unstable combustion, forced vibration of the rotor blades, damage to the engine structure integrity and even flameout and shutdown, seriously threatening flight safety.

[0004] Currently, surge detection usually collects compressor vibration data through an aviation accessory measurement device system. By performing frequency-domain analysis on the vibration data, the surge detection is realized by using the frequency-domain change characteristics. When using frequency-domain information for surge detection, it is often to perform surge detection by collecting data with a fixed time window. As the rotational speed of the aero-engine increases, the occurrence and development speed of the surge phenomenon may accelerate. Therefore, in order to detect surge in time and take measures to prevent its further deterioration, it is necessary to shorten the surge detection time. If the detection time is too long, it may lead to the situation that when the surge is detected, the surge phenomenon has developed to a relatively serious stage, thus causing greater harm. Summary of the Invention

[0005] In order to solve the technical problem of low timeliness in data collection and detection using a fixed time window, this application provides a data analysis method and system for an aviation accessory measurement device system.

[0006] In the first aspect, this application provides a data analysis method for an aviation accessory measurement device system, adopting the following technical solutions:

[0007] A data analysis method for an aviation accessory measurement device system, comprising the steps of: obtaining compressor surface vibration data within an initial time window, and obtaining a time-frequency diagram through a frequency-domain analysis algorithm; calculating a surge time window adaptation coefficient corresponding to the current time according to the time-frequency diagram, and adjusting the length of the initial time window according to the surge time window adaptation coefficient to obtain the optimal length of the next moment window; the calculation method of the surge time window adaptation coefficient is: obtaining the spectral curve corresponding to the maximum value of the time-domain projection of the time-frequency diagram, constructing a histogram of frequency values according to the spectral curve, and taking the frequency of the occurrence of the frequency value as the histogram frequency; calculating all the modes of the histogram frequency, and constructing a histogram frequency mode sequence; calculating an approximate mode, and the expression of the approximate mode is: ; represents the approximate mode; represents the current moment corresponding to the th degree of abnormality of the histogram frequency; represents the current moment corresponding to the minimum mode value in the histogram frequency mode sequence; represents a hyperparameter; represents the exponential function with as the base; taking the mode with the smallest Euclidean distance from the approximate mode as the optimal histogram frequency, constructing an optimal histogram frequency mode sequence, and calculating the information entropy value of the optimal histogram frequency mode sequence as the surge time window coefficient corresponding to the next moment.

[0008] The beneficial effects are as follows: Using the method of dynamic time window can significantly reduce the detection delay of abnormal signals compared with the traditional fixed time window. Under the high-speed operating conditions of the engine, this method can greatly improve the response speed of the system, effectively capture the precursor of instantaneous surge, so as to give early warning and take measures to avoid serious harm to the engine caused by surge. In addition, the adjustment of the dynamic time window is based on the surge time window adaptation coefficient, which is calculated by analyzing the frequency distribution characteristics in the time-frequency diagram, and can automatically adjust the length of the time window according to the actual change of the signal to ensure the best detection sensitivity and accuracy under different working conditions. At the same time, by optimizing the processing of histogram frequency and the calculation of information entropy value, the reliability of surge detection is further improved, and the possibility of false judgment and missed judgment is reduced.

[0009] Optionally, the degree of abnormality is the absolute difference between the histogram frequency and the mean value of the histogram frequency.

[0010] The beneficial effects are as follows: A method for quantifying the degree of abnormality is provided, which can comprehensively capture the degree of deviation of the frequency value from the normal level. Whether it is higher or lower than the mean value, it can be quantified as the degree of abnormality, so as to provide a more comprehensive reference for subsequent analysis.

[0011] Optionally, the calculation expression of the degree of abnormality is: ; Indicates the current moment The corresponding degree of abnormality of the histogram frequency; is the rectified linear unit function, Indicates the current moment The corresponding th histogram frequency; Indicates the mean value of the histogram frequency at the current moment .

[0012] The beneficial effect is: another method for quantifying the degree of abnormality is provided. By using the rectified linear unit function to process the difference between the histogram frequency and the mean value, high-amplitude abnormal signals can be highlighted while low-amplitude normal fluctuations are ignored. Abnormal signals that may cause surging can be accurately identified, so as to give early warnings and take measures to avoid potential safety risks.

[0013] Optionally, the optimal length of the current time window is obtained by adjusting the length of the initial time window according to the surge time window adaptation coefficient, including the steps of: performing binary classification on the optimal histogram frequency mode sequence to obtain a normal mode cluster and an abnormal mode cluster; calculating the optimal length; the expression of the optimal length is: ; In the formula, represents the optimal length corresponding to the moment Indicates the current moment the reciprocal of the minimum frequency value corresponding to the mode in the abnormal mode cluster; represents the first hyperparameter; represents the second hyperparameter; represents the exponential function with as the base; Indicates the current moment the corresponding surge time window adaptation coefficient.

[0014] The beneficial effect is: the reciprocal of the minimum frequency value in the abnormal mode cluster is used as a reference for the time window length, ensuring that the detection window can focus on the key abnormal signal areas. The first hyperparameter is used to adjust the multiple of the time window length to ensure the integrity of data acquisition; the second hyperparameter is used to control the influence degree of the surge time window adaptation coefficient on the window length. By reasonably adjusting these parameters, different engine operating conditions and detection requirements can be adapted, and the detection performance can be further optimized.

[0015] Optionally, the optimal length of the current time window is obtained by adjusting the length of the initial time window according to the surge time window adaptation coefficient, including the steps of: performing binary classification on the optimal histogram frequency mode sequence to obtain a normal mode cluster and an abnormal mode cluster; calculating the optimal length; the expression of the optimal length is: ; In the formula, represents The optimal length corresponding to the moment represents the current moment the reciprocal of the minimum frequency value corresponding to the mode in the abnormal mode cluster corresponding thereto; represents the second hyperparameter; represents the exponential function with as the base; represents the current moment the surge time window adaptation coefficient corresponding thereto.

[0016] The beneficial effect is that by adopting a simplified calculation method for the time window length, that is, without introducing an additional multiple adjustment parameter, the calculation complexity is reduced.

[0017] Optionally, perform binary classification on the optimal histogram frequency mode sequence to obtain a normal mode cluster and an abnormal mode cluster; calculate the optimal length; the expression of the optimal length is: ; in the formula, represents the optimal length corresponding to the moment represents the current moment the reciprocal of the minimum frequency value corresponding to the mode in the abnormal mode cluster corresponding thereto; represents the exponential function with as the base; represents the current moment the surge time window adaptation coefficient corresponding thereto.

[0018] Optionally, use the K-means clustering algorithm for binary classification.

[0019] Optionally, the frequency domain analysis algorithm is the music algorithm.

[0020] Optionally, the frequency domain analysis algorithm is the wavelet decomposition algorithm.

[0021] In a second aspect, the present application provides a data analysis system for an aviation accessory measurement device system, adopting the following technical solution:

[0022] A data analysis system for an aviation accessory measurement device system, comprising: a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the data analysis method for the aviation accessory measurement device system according to the above is implemented.

[0023] The beneficial effect is that the data analysis method for an aviation accessory measurement device system as described above is generated into a computer program and stored in the memory to be loaded and executed by the processor, thereby, making a system according to the memory and the processor is convenient to use.

[0024] The present application has the following technical effects:

[0025] 1. The method using a dynamic time window can significantly reduce the detection latency of abnormal signals compared to the traditional fixed time window. Under the condition of high-speed engine operation, this method can greatly improve the response speed of the system, effectively capture the precursors of instantaneous surge, and thus give early warnings and take measures to avoid serious damage to the engine caused by surge.

[0026] 2. The adjustment of the dynamic time window is based on the surge time window adaptation coefficient, which is calculated by analyzing the frequency distribution characteristics in the time-frequency diagram. This method can automatically adjust the length of the time window according to the actual changes of the signal, ensuring the best detection sensitivity and accuracy under different working conditions. At the same time, by optimizing the processing of the histogram frequency and the calculation of the information entropy value, the reliability of surge detection is further improved, and the possibilities of false positives and false negatives are reduced. Description of the Drawings

[0027] Figure 1 It is a flowchart of the method from step S1 to step S2 in the data analysis method of an aviation accessory measurement device system according to an embodiment of the present application.

[0028] Figure 2 It is a flowchart of the method of step S2 in the data analysis method of an aviation accessory measurement device system according to an embodiment of the present application. Detailed Embodiments

[0029] An embodiment of the present application discloses a data analysis method for an aviation accessory measurement device system. Referring to Figure 1 , it includes steps S1 - S2, which are specifically as follows:

[0030] S1: Obtain the vibration data on the surface of the compressor within the initial time window, and obtain the time-frequency diagram through the frequency domain analysis algorithm.

[0031] The aviation accessory measurement device system includes vibration sensors. A wide-band model (such as PCB 352C33, bandwidth 0.5Hz - 10kHz) is selected. The sensors of the aviation accessory measurement device system are placed at the intake section (low-pressure area) of the compressor, and 4 sensors are evenly arranged circumferentially with a phase difference of 90° to collect the vibration data of the aero-engine.

[0032] In one embodiment, the initial detection window length is set to Z = 2s. The collected data of the initial window is the vibration data collected within Z = 2s before the current moment. Specifically, assuming the current time is T, the data within the initial detection window refers to the vibration data collected from T - 2 seconds to T, that is, starting from the current moment, tracing back 2 seconds, and taking the vibration data within these 2 seconds as the data within the initial detection window.

[0033] Set the data collected in the initial window as , Indicates the current moment. For Use the MUSIC algorithm to obtain the current corresponding time-frequency diagram . The MUSIC algorithm is a spectrum estimation method mainly used for signal frequency analysis. By analyzing the covariance matrix of the signal, the signal space is divided into a signal subspace and a noise subspace, thereby achieving high-precision estimation of the signal frequency. The prior art will not be elaborated here.

[0034] In other embodiments, other frequency-domain analysis algorithms such as wavelet packet decomposition can also be used to obtain the time-frequency diagram . Wavelet packet decomposition is a multi-resolution analysis method based on wavelet transform, which can decompose the signal into sub-signals at different frequencies and time scales. Exemplarily, the Db4 wavelet basis function is adopted, and the decomposition level is set to 4 layers. The prior art will not be elaborated here.

[0035] S2: Calculate the surge time-window adaptation coefficient corresponding to the current time according to the time-frequency diagram, and adjust the length of the initial time window according to the surge time-window adaptation coefficient to obtain the optimal length of the next moment window.

[0036] After obtaining the time-frequency diagram , it is possible to perform surge detection in combination with the time-frequency diagram and other data such as air flow temperature. However, if surge detection is performed again at the next moment (for example seconds), the detection time interval is too long, which may lead to the inability to detect the surge phenomenon in time; and if the detection time interval is too short (for example seconds), it may not be possible to complete a complete surge detection process due to the short time.

[0037] Therefore, it is necessary to analyze the surge time-window adaptation coefficient of the current time-frequency diagram. Since the surge signal mainly shows low frequency and high amplitude, the present application sets the low-frequency threshold to 30 Hz, that is, only processes low-frequency data below 30 Hz (30 Hz is an adjustable hyperparameter, and the implementer can adjust it according to the specific scenario). The entropy value of the low-frequency information is calculated to represent the current abnormal degree of the surge. When the low-frequency amplitude increases, the data becomes more dispersed, and the corresponding entropy value also increases.

[0038] However, there is less high-amplitude frequency data generated during surge occurrence. Directly calculating the entropy value may lead to inaccurate results, thus affecting the judgment of surge abnormality. In order to avoid the interference of low-amplitude data on high-amplitude data, the present application preprocesses the data before calculating the entropy value, making the data form more suitable for using the entropy value as an index for surge abnormality detection, and at the same time can more accurately judge the rationality of the time window length.

[0039] The present application passes through the time-frequency diagram Preprocess the spectral curve, adjust the histogram frequency of the frequency values, which can highlight abnormal features and reduce interference. Calculate the adjusted entropy value as the surge time window adaptation coefficient. Specifically, the calculation method of the surge time window adaptation coefficient includes steps S20 - S22, as follows:

[0040] S20: Obtain the spectral curve corresponding to the maximum value of the time - domain projection of the time - frequency diagram, construct a histogram of the frequency values based on the spectral curve, and use the frequency of the frequency values as the histogram frequency.

[0041] After obtaining the time - frequency diagram, extract the spectral curve corresponding to the maximum value of the time - domain projection of the time - frequency diagram. The time - frequency diagram is a distribution diagram of the signal in time and frequency. Through time - domain projection, the frequency range corresponding to the maximum energy of the signal at a certain moment can be found. The spectral curve at the maximum value of the time - domain projection is selected because it reflects the main frequency characteristics of the signal at the current moment.

[0042] S21: Calculate all the modes of the histogram frequency and construct a sequence of histogram frequency modes.

[0043] Set the current moment as and extract the th frequency value from the spectral curve, as well as all the histogram frequencies corresponding to all the frequency values where the histogram frequency corresponding to the th frequency value is . Calculate the mean value of the histogram frequency and all the mode values where the th mode value is .

[0044] For each frequency value , calculate the absolute value of the difference between its histogram frequency value and the mean value and use as the degree of abnormality. The larger the value of the degree of abnormality, the higher the abnormal degree of the amplitude of this frequency value, and its abnormal features need to be retained without merging with the modes. The smaller the value of the degree of abnormality, the more it indicates that this frequency value belongs to normal amplitude fluctuations and needs to be merged to reduce interference. Calculate the difference between each frequency value on the spectral curve and the overall mean value, and distinguish the amplitude changes through these differences. By calculating the difference between each frequency value on the spectral curve and the overall mean value, the amplitude changes can be distinguished. High - amplitude frequency values are usually related to surges, while low - amplitude frequency values may be noise or normal signals.

[0045] Adjust the histogram frequency corresponding to each frequency value on the frequency spectrum curve to the mode of the closest histogram frequency value. This can reduce the interference of low-amplitude data on high-amplitude data, thereby increasing the variation range of the entropy value, enabling the entropy value to more accurately reflect the surge anomaly, and at the same time improving the effectiveness of subsequent time window division.

[0046] In other embodiments, the calculation expression for the degree of abnormality is: ; represents the current moment corresponding to the th histogram frequency's degree of abnormality; represents the current moment corresponding to the th histogram frequency; represents the mean value of the histogram frequency at the current moment . is the rectified linear function. When , ; when , .

[0047] When is greater than or equal to , = . If a certain frequency value is not less than the average frequency, its degree of abnormality is proportional to the excess part. At this time, the larger it is, the higher the degree of abnormality of the frequency. When is less than , , indicating that the situation below the average frequency will not be regarded as abnormal and the degree of abnormality is zero. When the histogram frequency is less than the mean value of the histogram frequency, the value of is 0, that is, the case of low amplitude is not considered; when the histogram frequency is greater than or equal to the mean value of the histogram frequency,

[0048] To adjust the frequency value to the most appropriate mode, define the local adjustment coefficient . Process the sequence of histogram frequency modes . If there are consecutive modes, discretize them by equidistant sampling (for example, the interval is 2). Obtain the maximum value and the minimum value in the sequence of mode values. is the th frequency value 's histogram frequency value The local adjustment coefficient corresponding to Adjust to and The most appropriate mode is determined to reduce the interference of low-frequency and low-amplitude vibrations and increase the difference of low-frequency and high-amplitude vibrations.

[0049] The specific acquisition process is: Get the histogram frequency mode sequence Middle Mode ,exist There may be a situation where the majority is continuous. If the majority is continuous, the continuous majority is sampled at equal intervals to obtain a discrete majority sequence, where the continuous judgment method is achieved through a counting method, which is not repeated in this scheme. The interval length of the equal interval sampling is 2, which can be adjusted by the implementer according to the specific implementation scenario.

[0050] In one embodiment, in order to avoid the extreme situation where the mode is too small, if the mode is lower than a preset number (5), it is defaulted to detect sensor failure.

[0051] S22: Calculate the approximate mode; take the mode with the smallest Euclidean distance to the approximate mode as the optimal histogram frequency, construct the optimal histogram frequency mode sequence, calculate the information entropy value of the optimal histogram frequency mode sequence and use it as the surge time window coefficient of the corresponding window at the next moment.

[0052] The expression for the approximate mode is: ; represents the approximate mode; Indicates the current time The corresponding The degree of abnormality of the histogram frequency; Indicates the current time The minimum mode value in the corresponding histogram frequency mode sequence; Indicates An exponential function with base . represents the third hyperparameter, which can be 1 for example. The value can be adjusted according to the actual application scenario.

[0053] The degree of abnormality Negative correlation mapping, the mapped value and the minimum mode Multiply them together to get the approximate mode, which is an adjusted frequency value. When is smaller (i.e. the frequency value is close to the normal level), the adjusted frequency value is closer to the mode, thus achieving merging and unification. When it is large (i.e., the degree of abnormality of the frequency value is high), the adjusted frequency value differs greatly from the original value, and merging is not carried out as much as possible.

[0054] Construct an optimal histogram frequency mode sequence based on the adjusted frequency value (optimal histogram frequency), calculate the information entropy value for the obtained new sequence, and use it as the surge time window adaptation coefficient for the corresponding window at the current moment.

[0055] The larger the surge time window adaptation coefficient, the more chaotic the low-frequency frequency is, and thus the greater the possibility of surge. On the contrary, the possibility of surge is smaller. In order to detect surge anomalies in a timely manner, the time window should be narrowed and the detection frequency should be increased. However, the time window cannot be too small so as not to lose too much information, and the surge time window adaptation coefficient cannot be too small to prevent excessive loss of information. Therefore, the present application sets the minimum window length value to 0.5 s, and 0.5 s can be adjusted by the implementer according to the specific implementation scenario. The reason for not directly using the minimum window length value is that if the minimum window length value is directly used, more low-frequency data may be lost, and the time window needs to be adjusted according to the current frequency domain distribution.

[0056] Adjust the length of the initial time window according to the surge time window adaptation coefficient to obtain the optimal length of the window at the next moment. The calculation method of the optimal length is as follows:

[0057] Perform binary classification on the optimal histogram frequency mode sequence to obtain a normal mode cluster and an abnormal mode cluster; the modes in the optimal histogram frequency mode sequence should be divided into two parts. The part with a small difference from the original data mean represents the normal mode, and the part with a large difference from the original data mean represents the abnormal mode. Furthermore, after obtaining the new adjusted frequency value sequence, perform binary classification analysis on these frequency values to distinguish normal and abnormal modes.

[0058] Calculate the optimal length; the expression of the optimal length is: ; where represents the optimal length corresponding to the moment, represents the reciprocal of the minimum frequency value corresponding to the mode in the abnormal mode cluster corresponding to the current moment represents the exponential function with as the base; represents the surge time window adaptation coefficient corresponding to the current moment

[0059] In order to make the calculation of the optimal length adaptable to more scenarios, the expression of the optimal length can be: ; where represents the optimal length corresponding to the moment, The reciprocal of the minimum frequency value corresponding to the mode in the corresponding abnormal mode cluster; Denote the exponential function with as the base; Denote the surge window adaptation coefficient corresponding to the current moment ;

[0060] Wherein, Denote the first hyperparameter; Denote the second hyperparameter; Exemplarily, the first hyperparameter can be 5 and the second hyperparameter can be 1.5; The first hyperparameter and the second hyperparameter can be adjusted or not used according to the actual application scenario in the calculation of the optimal length, which will not be elaborated here.

[0061] In one embodiment, the K-means clustering algorithm is used to divide the optimal histogram frequency mode sequence into two categories. Each mode in each category corresponds to a frequency value, and the abnormality degree (denoted as value) of these frequency values is used to distinguish normal and abnormal modes. Calculate the average value (denoted as value) of the values of each mode in each category. Then, the average value of all values in each category is calculated again to obtain the overall average value (denoted as value) of this category. Finally, compare the values of the two categories. Among them, The category with a larger value is determined as the abnormal mode category because a larger value indicates that the frequency values in this category have a higher degree of abnormality.

[0062] Due to the inability to perform real-time sampling, this solution selects the start time point of the next time window for surge detection. To ensure the continuity and coverage of the detection, the selected time point is the time corresponding to a 30% overlap rate after the end of the current time window. This can ensure the coherence of the detection process and avoid missing key information due to too long a time interval.

[0063] The embodiment of the present application also discloses a data analysis system for an aviation accessory measurement device system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the data analysis method for the aviation accessory measurement device system according to the present application is implemented. The above system also includes other components well known to those skilled in the art such as a communication bus and a communication interface, and their settings and functions are known in the art, so they will not be elaborated here.

[0064] In this application, the aforementioned memory can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, or any other medium that can be used to store the required information and can be accessed by an application program, module, or both.

[0065] The above are all the preferred embodiments of this application, and the protection scope of this application is not limited thereby. Therefore, any equivalent changes made according to the structure, shape, and principle of this application shall be covered within the protection scope of this application.

Claims

1. A data analysis method for an aviation accessory measuring device system, characterized in that: The method comprises the following steps: obtaining compressor surface vibration data within an initial time window, obtaining a time-frequency diagram through a frequency domain analysis algorithm; calculating a surge time window adaptation coefficient corresponding to the current time according to the time-frequency diagram, and adjusting the length of the initial time window according to the surge time window adaptation coefficient to obtain an optimal length of the window at the next moment; The calculation method of the surge time window adaptation coefficient is as follows: obtain the spectrum curve corresponding to the maximum value of the time-frequency diagram at the time domain projection, construct a histogram of the frequency value according to the spectrum curve, and take the frequency of occurrence of the frequency value as the histogram frequency; Calculate all modes of the histogram frequency and construct a histogram frequency mode sequence; calculate the approximate mode, and the expression of the approximate mode is: ; represents the approximate mode; Indicates the current time The corresponding The degree of abnormality of the histogram frequency; Indicates the current time The minimum mode value in the corresponding histogram frequency mode sequence; represents a hyperparameter; Indicates An exponential function with the base ; the mode with the smallest Euclidean distance to the approximate mode is taken as the optimal histogram frequency, the optimal histogram frequency mode sequence is constructed, and the information entropy value of the optimal histogram frequency mode sequence is calculated and used as the surge time window coefficient of the corresponding window at the next moment.

2. The data analysis method of the aviation accessory measuring device system according to claim 1, characterized in that: The degree of abnormality is the absolute difference between the histogram frequency and the mean of the histogram frequency.

3. The data analysis method of the aviation accessory measuring device system according to claim 1, characterized in that: The calculation expression of abnormality degree is: ; Indicates the current time The corresponding The degree of abnormality of the histogram frequency; is a linear rectification function, Indicates the current time The corresponding Histogram frequencies; Indicates the current time The histogram frequency mean of .

4. The data analysis method for the aviation accessory measuring device system according to claim 1, characterized in that: The length of the initial time window is adjusted according to the surge time window adaptation coefficient to obtain the optimal length of the current time window, including the following steps: The optimal histogram frequency mode sequence is classified into two categories to obtain normal mode clusters and abnormal mode clusters; the optimal length is calculated; the expression of the optimal length is: ; In the formula, express The optimal length corresponding to the moment, Indicates the current time The reciprocal of the minimum frequency value corresponding to the mode in the corresponding abnormal mode cluster; represents the first hyperparameter; represents the second hyperparameter; Indicates An exponential function with base ; Indicates the current time The corresponding surge time window adaptation coefficient.

5. The data analysis method for the aviation accessory measuring device system according to claim 1, characterized in that: The length of the initial time window is adjusted according to the surge time window adaptation coefficient to obtain the optimal length of the current time window, including the following steps: The optimal histogram frequency mode sequence is classified into two categories to obtain normal mode clusters and abnormal mode clusters; the optimal length is calculated; the expression of the optimal length is: ; In the formula, express The optimal length corresponding to the moment, Indicates the current time The reciprocal of the minimum frequency value corresponding to the mode in the corresponding abnormal mode cluster; represents the second hyperparameter; Indicates An exponential function with base ; Indicates the current time The corresponding surge time window adaptation coefficient.

6. The data analysis method for the aviation accessory measuring device system according to claim 1, characterized in that: The length of the initial time window is adjusted according to the surge time window adaptation coefficient to obtain the optimal length of the current time window, including the following steps: The optimal histogram frequency mode sequence is classified into two categories to obtain normal mode clusters and abnormal mode clusters; the optimal length is calculated; the expression of the optimal length is: ; In the formula, express The optimal length corresponding to the moment, Indicates the current time The reciprocal of the minimum frequency value corresponding to the mode in the corresponding abnormal mode cluster; Indicates An exponential function with base ; Indicates the current time The corresponding surge time window adaptation coefficient.

7. The data analysis method of the aviation accessory measuring device system according to any one of claims 4 to 6, characterized in that: The K-means clustering algorithm is used for binary classification.

8. The data analysis method for the aviation accessory measuring device system according to claim 1, characterized in that: The frequency domain analysis algorithm is the music algorithm.

9. The data analysis method for the aviation accessory measuring device system according to claim 1, characterized in that: The frequency domain analysis algorithm is wavelet decomposition algorithm.

10. A data analysis system for an aviation accessory measuring device system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a data analysis method for an aviation accessory measuring device system according to any one of claims 1 to 9 is implemented.

Images