Aircraft state monitoring method and system based on big data

Abstract

The invention relates to the technical field of data processing, in particular to an aircraft state monitoring method and system based on big data, and the method comprises the steps: obtaining the operation data of an aircraft, constructing an initial Markov chain, constructing a graph data structure of each numerical interval according to each operation state in the initial Markov chain, and obtaining a graph data structure of each numerical interval; and the numerical intervals are optimized by comparing the graph data structures of the numerical intervals, so that an optimal Markov chain is constructed, and the running state of the aircraft is monitored based on the optimal Markov chain. The method has the effect of improving the monitoring accuracy of the running state of the aircraft.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular, to an aircraft status monitoring method and system based on big data. Background Art

[0002] An aircraft refers to a flying object that can fly off the ground and be used by people, which is of great significance to multiple industries. For example, in the transportation field, aircraft shorten the distance between people and the distance between objects; in the crop planting industry, aircraft can replace manual spraying. In addition, the application of aircraft can be seen in fields such as surveying and mapping, monitoring, and aviation. However, with the development of aircraft fly-by-wire flight control and onboard networks, the number of aircraft health monitoring sensors has been continuously increasing, and the real-time data during aircraft flight shows characteristics such as multi-dimensional, high-frequency, and long time series. There are certain limitations in the process of monitoring the status of an aircraft using traditional fault tree or correlation graph analysis methods. Therefore, the operation status monitoring of an aircraft usually uses a Markov model for monitoring. The Markov model can start from the state of the system, analyze the connection between the system and the state, and then put forward requirements for the failure probability of the system.

[0003] However, there are also certain limitations in the process of using the traditional Markov model to monitor the operation status of an aircraft. In the traditional Markov model, each state is divided and fixed manually. When facing different flight segments, environments, and aging conditions, state sparsity, boundary distortion, and model degradation are likely to occur, making it difficult to maintain a stable transition probability while maintaining high resolution, resulting in frequent missed detections and false alarms. Summary of the Invention

[0004] In order to solve the problem of frequent missed detections and false alarms in the traditional monitoring method of aircraft, this application provides an aircraft status monitoring method and system based on big data.

[0005] In a first aspect, this application provides an aircraft status monitoring method based on big data, adopting the following technical solution: The aircraft status monitoring method based on big data includes the steps of: collecting data during the operation of the aircraft, segmenting the operation data to form multiple standard segments; for the multi-dimensional data in any standard segment, obtaining the value range of each dimension data, uniformly dividing the value range to form multiple numerical intervals; different combinations of numerical intervals in different dimensions constitute an operation state, and multiple operation states constitute the initial state space of the standard segment, and statistically calculating the transition probability between different operation states to construct an initial Markov chain; Obtain the graph data structure for each numerical interval; calculate the similarity of the graph data structures of adjacent numerical intervals, and merge or divide the adjacent numerical intervals based on the similarity of the graph data structures to construct the optimal interval; construct the optimal state space based on the optimal interval, and based on the optimal state space, statistically calculate the transition probabilities between each operating state in each optimal state space to construct the optimal Markov chain, and monitor the flight state of the aircraft according to the optimal Markov chain.

[0006] The beneficial effects are as follows: Construct the initial state space based on the operating data of the aircraft. Then construct the graph data structure corresponding to the numerical interval of each dimension data in the initial space, and merge or divide the numerical intervals according to the similarity between the graph data structures. During the process of dividing and merging the numerical space, if the similarity between the graph data structures corresponding to two numerical intervals is greater, it indicates that the transition probabilities of the operating states corresponding to these two intervals are approximate, indicating that the two intervals can be merged to reduce the redundant operating states in the state space and simplify the calculation steps and monitoring model. If the graph data structures of two numerical intervals are extremely small, it indicates that the transition probabilities of the operating states corresponding to adjacent numerical intervals differ greatly, which may be caused by insufficiently fine division of the numerical intervals. Therefore, the numerical intervals are divided again to improve the accuracy and precision of the final monitoring.

[0007] At the same time, in this method, division and calculation are performed based on the data of the aircraft operation, and at the same time, the data during the operation of the aircraft is continuously obtained according to satellite communication to perform rolling update on the large dataset containing the flight state of the aircraft. Moreover, the division of the optimal interval is also updated dynamically. Compared with the fixed state boundary in the related technology, the method in this application can adapt to more different flight segments and more complex environments, reducing the occurrence of missed detections or false alarms.

[0008] Optionally, use the graph kernel method to calculate the similarity of the graph data structures.

[0009] The beneficial effects are as follows: The graph kernel method is a technology for graph data classification and similarity measurement. Through this method, the similarity between the graph data structures corresponding to numerical intervals can be measured.

[0010] Optionally, the steps of merging or dividing adjacent numerical intervals based on the similarity of the graph data structures to construct the optimal interval include: for multiple similarities corresponding to multiple numerical intervals, the multiple similarities form an approximate sequence, and use an outlier detection algorithm to detect the outliers in the approximate sequence; obtain the mean value of all similarities; in response to the outlier being greater than the mean value of the similarities, merge the two numerical intervals corresponding to the outlier; in response to the outlier being less than the mean value of the similarities, divide the numerical interval corresponding to the outlier until the division stop condition is met, and complete the construction of the optimal interval.

[0011] The beneficial effects are as follows: Obtain the similarity of two adjacent numerical intervals to form an approximate sequence. Detect abnormal data in the approximate sequence through an anomaly detection algorithm, so as to obtain some abnormally small or large data in the approximate sequence, and then be able to judge that the two numerical intervals are too similar or too different, thus facilitating the subsequent merging or partitioning of the numerical intervals.

[0012] Optionally, the steps of merging or partitioning adjacent numerical intervals based on the similarity of the graph data structure to construct the optimal interval include: For multiple similarities corresponding to multiple numerical intervals, the multiple similarities form an approximate sequence, use the least squares method to interpolate the data to form an optimized sequence; Use an anomaly detection algorithm to detect the outliers in the optimized sequence; Obtain the mean value of all similarities; In response to the outlier being greater than the mean value of the similarities, merge the two numerical intervals corresponding to the outlier; In response to the outlier being less than the mean value of the similarities, partition the numerical interval corresponding to the outlier until the partitioning stop condition is met, and complete the construction of the optimal interval.

[0013] The beneficial effects are as follows: In this method, the difference method is used to expand the data volume in the approximate sequence to improve the accuracy of subsequent anomaly detection by the anomaly detection algorithm, and further improve the accuracy of numerical interval partitioning.

[0014] Optionally, the method of partitioning the numerical interval corresponding to the outlier includes: For any interval that needs to be partitioned again, divide it into two equal numerical intervals.

[0015] The beneficial effects are as follows: Uniformly divide the numerical interval to be partitioned, so as to quickly complete the re-partitioning of the numerical interval and improve the calculation efficiency.

[0016] Optionally, the partitioning stop condition is: The distance between the optimized sequences corresponding to the numerical intervals before and after partitioning is less than a preset threshold or the preset partitioning times are reached.

[0017] The beneficial effects are as follows: Set a stop condition for the partitioning of the numerical interval to avoid endless partitioning of the numerical interval and improve the calculation efficiency.

[0018] Optionally, the distance between the optimized sequences corresponding to the numerical intervals before and after partitioning is the DTW distance between the two optimized sequences.

[0019] The beneficial effects are as follows: Reflect the difference between the two approximate sequences through the DTW distance of the approximate sequences corresponding before and after partitioning, and then judge whether the numerical interval needs to be further partitioned. If the DTW distance value is larger, it means that the transition probability of the operating state in the state space of the numerical interval before and after partitioning has changed greatly, so it needs to be partitioned again.

[0020] Optionally, the steps of constructing an optimal state space based on the divided intervals include: Any two optimal intervals in different dimensions form an optimal operating state, and multiple optimal operating states form an optimal state space.

[0021] Optionally, the steps of constructing an optimal state space based on the divided intervals include: For each dimension in each standard segment, obtain the numerical points of the numerical interval division, set a numerical window that can slide, and count the number of numerical points located in the numerical window during the sliding of the numerical window to construct a density curve. Take the peak point of the density curve as the optimal division point, and divide the corresponding dimension data based on the optimal division point to form a fusion interval. Any two fusion intervals in different dimensions form a fusion state, and multiple fusion states form an optimal state space.

[0022] The beneficial effect is that: A single standard segment is for local data, while in this method, by fusing the optimal intervals of multiple standard segments to form a fusion interval, the situation of uneven interval division in a single standard segment is reduced, and the accuracy and robustness of the final monitoring are improved.

[0023] In a second aspect, the present application provides a flight vehicle state monitoring system based on big data, adopting the following technical solution: The flight vehicle state monitoring system based on big data includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the flight vehicle state monitoring method based on big data as described above is implemented.

[0024] The beneficial effect is that: The flight vehicle state monitoring method based on big data as described above is generated into a computer program and stored in the memory to be loaded and executed by the processor. Thus, a system is made according to the memory and the processor, which is convenient to use.

[0025] The present application has the following technical effects: 1. Dynamically divide numerical intervals based on historical data, construct an optimal Markov chain, and improve the accuracy of monitoring the operating state of the flight vehicle.

[0026] 2. Merge similar numerical intervals, reduce redundant states, simplify the monitoring steps, and improve the calculation efficiency.

[0027] 3. Flight data is uploaded to the cloud server through satellite communication, and the cloud server monitors the flight state of the flight vehicle based on big data, supporting parallel detection of a large-scale flight vehicle cluster and ensuring the efficiency of data processing.

[0028] 4. The optimal Markov chain constructed based on big data can be automatically optimized according to the update and iteration of big data, adapting to long-term slow evolution trends such as flight vehicle aging and environmental changes. Description of the Drawings

[0029] Figure 1 It is the flowchart of the method for monitoring the state of an aircraft based on big data according to an embodiment of the present application.

[0030] Figure 2 It is the density curve graph of the method for monitoring the state of an aircraft based on big data according to an embodiment of the present application. Detailed implementation manners

[0031] The embodiment of the present application discloses a method for monitoring the state of an aircraft based on big data, which acquires the operation data of the aircraft, segments the data to form multiple standard segments. For each standard segment, which includes data in multiple dimensions, for any data in one dimension, it is divided into multiple numerical intervals according to the value range of its value, and an initial state space and a Markov chain are constructed according to the numerical intervals; according to the graph data structure of each operating state in the state space, the numerical intervals are adaptively subdivided and merged to adjust the state boundaries. While ensuring the resolution, the redundant states in the state space are compressed to construct an optimal state space, improving the accuracy of anomaly detection. At the same time, the numerical intervals can be dynamically divided according to the online rolling update of big data, facilitating the long-term health management of the operating state of the aircraft.

[0032] Referring to Figure 1 , the method for monitoring the state of an aircraft based on big data includes: step S1 - step S2.

[0033] S1: Collect data during the operation of the aircraft, segment the operation data to form multiple standard segments; for the multi-dimensional data in any standard segment, obtain the value range of each dimension data, and evenly divide the value range to form multiple numerical intervals.

[0034] During the flight of the aircraft, the flight data of the engine is collected by on-board sensors and compressed into data packets. The data packets are uploaded to the cloud server through the satellite communication module, and the cloud server processes the flight data. The flight data includes data in multiple dimensions, such as: flight speed, flight acceleration, engine temperature, engine speed, etc. The normal flight data of the aircraft is segmented in multiple segments according to time series, and when segmenting, the normal flight data of the aircraft is segmented with a fixed time length. In this embodiment, the segment duration is set to 3 days of flight duration. In other embodiments, other lengths of segment durations can be used.

[0035] After the flight data is segmented, it forms multiple standard segments. For any standard segment, it includes data information in multiple dimensions. It can also be understood that during the operation of the aircraft, information in multiple dimensions of the aircraft is collected simultaneously.

[0036] For any data in one dimension, obtain its value range, and evenly divide the value range to make it form multiple numerical intervals.

[0037] S2: Different combinations of numerical value ranges in different dimensions constitute an operating state, and multiple operating states constitute the initial state space of this standard section. The transition probabilities between different operating states are statistically calculated to construct an initial Markov chain.

[0038] After the numerical value ranges are divided, for any one dimension, there are multiple corresponding numerical value ranges. The numerical value ranges of different dimensions are combined to constitute the operating states of the aircraft, and different operating states constitute the state space of the aircraft's operation. The transition probabilities between different operating states in the state space are statistically calculated to construct an initial Markov chain.

[0039] Exemplarily, suppose in this embodiment, data of three dimensions of the aircraft are collected, namely engine temperature, engine speed, and aircraft flight speed; the value range of the engine temperature is 0 - 900 ; the value range of the engine speed is: 0 - 8000 r / min; the value range of the aircraft flight speed is: 0 - 300 km / h. The value ranges of the data of each dimension are divided into three numerical value ranges. For the convenience of description, here the multiple numerical value ranges corresponding to the engine temperature are respectively defined as the low-temperature range, normal-temperature range, and high-temperature range; the multiple numerical value ranges corresponding to the engine speed are respectively defined as the low-speed range, normal-speed range, and high-speed range; the multiple numerical value ranges corresponding to the aircraft flight speed are respectively defined as the low-speed range, medium-speed range, and high-speed range. Then an exemplary operating state of the aircraft is: the operating state corresponding to the combination of the normal-temperature range, normal-speed range, and medium-speed range. And the set of operating states formed by any combination of the above ranges is the state space of the aircraft.

[0040] In this embodiment, in order to improve the accuracy of anomaly detection, the value ranges corresponding to the data of each dimension in the standard section are divided into 10 numerical value ranges.

[0041] S3: Obtain the graph data structure of each numerical value range; For a numerical value range corresponding to the data of any one dimension, it corresponds to multiple operating states, and this type of state is defined as an interval state. In the state space constructed in the above step S1, there are transition probabilities between different operating states. Obtain the operating states that have transition probabilities with the interval states, and define this type of operating state as the connection space. Use the connection space and the operating states as the nodes of the graph data structure, and use the transition probabilities between different nodes as the edge weights between the two nodes of the graph data structure to complete the construction of the graph data structure corresponding to this interval.

[0042] S4: Calculate the similarity of the graph data structures of adjacent numerical value ranges, and based on the similarity of the graph data structures, merge or divide adjacent numerical value ranges to construct the optimal range.

[0043] In this embodiment, the similarity of the graph data structures corresponding to two adjacent intervals is calculated by the graph kernel method. The graph kernel method is a conventional technical means in the art and will not be elaborated here.

[0044] If there are approximate graph data structures in different numerical intervals of the same dimension, it indicates that there are numerical intervals that can be merged. If there are significantly different graph data structures in different numerical intervals of the same dimension, it indicates that the division of the numerical intervals is not fine enough.

[0045] For the combination of different numerical intervals in different dimensions, if two numerical intervals are adjacent and their graph data structures are relatively similar, then it is considered that the two numerical intervals need to be merged. If two numerical intervals are far apart, then the influence between them is small and can be ignored. Therefore, in this embodiment, only the approximation of two adjacent numerical intervals is calculated to simplify the calculation steps and improve the calculation efficiency.

[0046] For multiple similarities corresponding to multiple numerical intervals, the multiple similarities form an approximate sequence, and an outlier detection algorithm is used to detect outliers in the approximate sequence; the mean value of all similarities is obtained; in response to the outlier being greater than the mean value of the similarities, the two numerical intervals corresponding to the outlier are merged; in response to the outlier being less than the mean value of the similarities, the numerical interval corresponding to the outlier is divided until the division stop condition is met, and the construction of the optimal interval is completed.

[0047] For any dimension, after its numerical intervals are arranged from large to small, then the similarity of the graph data structures between each numerical interval and its adjacent numerical interval is calculated. The multiple similarities form an approximate sequence.

[0048] In one embodiment, outliers in the approximate sequence are directly detected by an outlier detection algorithm. After the outliers are detected, the outliers are extracted and the two numerical intervals corresponding to the outliers are obtained. The outlier is compared with the mean value of the data in the approximate sequence. If the outlier is less than the mean value of the data in the approximate sequence, it indicates that the approximation degree of the two numerical intervals is too small and needs to be divided again. Here, the outlier detection algorithm can be the three - standard - deviation method or the quartile method, and such methods are conventional technical means in the art and will not be elaborated here.

[0049] If the outlier is greater than the mean value of the data in the approximate sequence, it indicates that the approximation degree of the two numerical intervals is too large, and the two numerical intervals need to be merged, thereby reducing the redundant operating states in the state space.

[0050] In another embodiment, the least squares method can be used to interpolate the data in the approximate sequence, and an optimized sequence is obtained after interpolating the approximate sequence. The optimized sequence is subjected to anomaly detection through an anomaly detection algorithm, and then the anomaly values are obtained to judge the division or combination of two intervals. The data volume is expanded by the difference method to reduce the error in the process of interval division or combination.

[0051] For any two numerical intervals to be combined, the combination method is: the two numerical intervals are directly combined into one numerical interval.

[0052] For any two numerical intervals to be divided, the division method is to divide a numerical interval into two equal parts. In other embodiments, it can also be divided into three equal parts.

[0053] In some special cases, a numerical interval may not only need to be combined, but also need to be divided. For example, a numerical interval needs to be combined with the adjacent numerical interval on the left, while the adjacent numerical interval on the right needs to be divided. For such cases, the method of dividing first and then combining is adopted, that is, the numerical interval is first divided, and then combined based on the divided numerical intervals.

[0054] After the numerical interval is divided or combined, which can also be understood as the numerical interval is optimized once here, the initial state space is reconstructed. And the graph data structure corresponding to each numerical interval and the similarity between each graph data structure are obtained, and multiple similarities also form an approximate sequence again. Calculate the DTW distance between the approximate sequence corresponding before optimization and the approximate sequence corresponding after optimization, and judge whether the numerical interval needs to be further optimized according to the DTW distance.

[0055] In this embodiment, the DTW threshold is set to 3.4. In response to the DTW distance between the approximate sequence before and after optimization or the optimized sequence being less than the DTW threshold or the number of times of numerical interval division reaching the preset number of divisions, stop the anomaly detection of the approximate sequence and the division and combination of the numerical interval. And take the numerical interval obtained by the last division as the optimal interval. In this embodiment, the preset number of divisions is set to 10, and in other embodiments, it can be adjusted according to the actual situation.

[0056] S5: Construct an optimal state space based on the optimal interval, statistically calculate the transition probability between each operating state in each optimal state space based on the optimal state space, construct an optimal Markov chain, and monitor the flight state of the aircraft according to the optimal Markov chain.

[0057] In one embodiment, the steps of constructing an optimal state space based on the divided intervals include: after obtaining the optimal intervals of each dimension of a standard segment, the optimal intervals of each dimension of the standard segment are combined with each other to form an optimal state space, and the transition probabilities between the operating states in the optimal state space are statistically calculated to construct an optimal Markov chain.

[0058] In another embodiment, the steps of constructing an optimal state space based on the divided intervals include: calculating the optimal intervals of each standard segment, and then fusing the optimal intervals of different standard segments.

[0059] The fusion method of the optimal intervals of different standard segments includes the steps of: for each dimension in each standard segment, obtaining the numerical points of the numerical interval division. Setting a slidable numerical window, and statistically calculating the number of numerical points located in the numerical window during the sliding process of the numerical window to construct a density curve, taking the peak point of the density curve as the optimal division point, dividing the corresponding dimension data based on the optimal division point to form a fusion interval, and any two fusion intervals of different dimensions form a fusion state, and multiple fusion states form an optimal state space.

[0060] Each standard segment includes data of the same dimension. For example, each standard segment includes engine temperature data, engine speed data, and flight speed data. The optimal intervals of the same dimension in different standard segments are different. Therefore, the optimal spaces of different dimensions in different standard segments are fused to improve the accuracy of the final monitoring of the flight state of the aircraft.

[0061] Set a slidable numerical window, and statistically calculate the number of numerical points that divide the optimal window in the numerical window to generate a density curve. Here, the numerical points refer to the data points used to divide the optimal window. For example, if the optimal interval of a dimension is divided into , , , then the numerical points corresponding to the optimal interval of this dimension are 0, 100, 200, and 300 respectively.

[0062] The numerical points of the optimal intervals of the same dimension in different standard segments are not consistent. Therefore, a slidable numerical window is set, and the number of numerical points in the numerical window is statistically calculated. Each peak point corresponds to a numerical window. For example, the interval corresponding to the numerical window corresponding to a peak point is , and the number of numerical points corresponding to this peak point is 3, which means that in this numerical window, the numerical points of the optimal intervals of three standard segments are located in the interval .

[0063] Refer to Figure 2, generate a density curve according to the number of numerical points, and perform data smoothing on the initial density curve using a data smoothing method. Specifically, a coordinate system can be constructed with the horizontal axis as the number axis and the vertical axis representing the number of numerical points. At the beginning of the sliding window sliding, the left endpoint of the numerical window is 0, and the right endpoint corresponds to the length of the numerical window. If there are three numerical points in this numerical window, then the coordinates of the point corresponding to this numerical window in the density curve are (0, 3); slide the numerical window and count the number of numerical points at different positions of the numerical window to form a density curve.

[0064] Perform peak point detection based on the smoothed density curve, divide the value range of each dimension according to the position corresponding to the peak point to form a fusion interval, and construct an optimal state space according to the combination of the fusion intervals of different dimensions. The construction of the optimal state space is the same as that of the initial state space and will not be elaborated here.

[0065] Collect the real-time operation state data during the flight of the aircraft, use the optimal Markov chain for prediction, compare the measured result with the prediction result of the Markov chain. When the measured operation data is not within the optimal interval corresponding to the operation state predicted by the Markov chain, it indicates that the aircraft is operating abnormally, and the monitoring of the aircraft operation state is realized.

[0066] In this method, based on satellite communication, the large dataset during the flight of the aircraft is updated in real time, and the optimal Markov chain is updated and optimized in real time, which can not only synchronously monitor the aircraft cluster, but also improve the accuracy of monitoring the flight state of the aircraft.

[0067] The embodiment of the present application also discloses a flight vehicle state monitoring system based on big data, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the flight vehicle state monitoring method based on big data of the present application is realized.

[0068] 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.

[0069] The above are all the preferred embodiments of the present application. The protection scope of the present application is not limited by this, so: all equivalent changes made according to the structure, shape, and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method for monitoring the state of an aircraft based on big data, characterized in that It includes the steps of: collecting data during the operation of the aircraft, segmenting the operation data to form multiple standard segments; for the multi-dimensional data in any standard segment, obtaining the value ranges of the data in each dimension, evenly dividing the value ranges to form multiple numerical intervals; different combinations of numerical intervals in different dimensions constitute an operation state, and multiple operation states constitute the initial state space of the standard segment, and statistically calculating the transition probabilities between different operation states to construct an initial Markov chain; Obtaining the graph data structure of each numerical interval; calculating the similarity of the graph data structures of adjacent numerical intervals, and merging or dividing adjacent numerical intervals based on the similarity of the graph data structures to construct optimal intervals; Constructing an optimal state space based on the optimal intervals, and statistically calculating the transition probabilities between the operation states in each optimal state space based on the optimal state space to construct an optimal Markov chain, and monitoring the flight state of the aircraft according to the optimal Markov chain.

2. The method for monitoring the state of an aircraft based on big data according to claim 1, wherein Using the graph kernel method to calculate the similarity of the graph data structures.

3. The method for monitoring the state of an aircraft based on big data according to claim 1, characterized in that The steps of merging or dividing adjacent numerical intervals based on the similarity of the graph data structures to construct optimal intervals include: for multiple similarities corresponding to multiple numerical intervals, the multiple similarities form an approximate sequence, and using an anomaly detection algorithm to detect the outliers in the approximate sequence; obtaining the mean value of all similarities; in response to the outlier being greater than the mean value of the similarities, merging the two numerical intervals corresponding to the outlier; in response to the outlier being less than the mean value of the similarities, dividing the numerical interval corresponding to the outlier until the division stop condition is met, and completing the construction of the optimal intervals.

4. The method for monitoring the state of an aircraft based on big data according to claim 1, characterized in that, The steps of merging or dividing adjacent numerical intervals based on the similarity of the graph data structures to construct optimal intervals include: for multiple similarities corresponding to multiple numerical intervals, the multiple similarities form an approximate sequence, using the least squares method to interpolate the data to form an optimized sequence; using an anomaly detection algorithm to detect the outliers in the optimized sequence; obtaining the mean value of all similarities; in response to the outlier being greater than the mean value of the similarities, merging the two numerical intervals corresponding to the outlier; in response to the outlier being less than the mean value of the similarities, dividing the numerical interval corresponding to the outlier until the division stop condition is met, and completing the construction of the optimal intervals.

5. The method for monitoring the state of an aircraft based on big data according to claim 3 or 4, characterized in that, The method of dividing the numerical interval corresponding to the outlier includes: for any interval that needs to be divided again, equally dividing it into two numerical intervals.

6. The method for monitoring the state of an aircraft based on big data according to claim 4, characterized in that The division stop condition is: the distance between the optimized sequences corresponding to the numerical intervals before and after division is less than a preset threshold or the preset division times are reached.

7. The method for monitoring the state of an aircraft based on big data according to claim 6, wherein The distance between the optimized sequences corresponding to the numerical intervals before and after division is the DTW distance between the two optimized sequences.

8. The method for monitoring the state of an aircraft based on big data according to claim 7, wherein, The steps of constructing an optimal state space based on the divided intervals include: any two optimal intervals in different dimensions constitute an optimal operation state, and multiple optimal operation states constitute an optimal state space.

9. The method for monitoring the state of an aircraft based on big data according to claim 1, characterized in that, The steps of constructing an optimal state space based on the divided intervals include: for each dimension in each standard segment, obtain the numerical points of the numerical interval division, set a slidable numerical window, and during the sliding process of the numerical window, count the number of numerical points located in the numerical window to construct a density curve, take the peak point of the density curve as the optimal division point, divide the corresponding dimension data based on the optimal division point to form a fusion interval, and any two fusion intervals of different dimensions constitute a fusion state, and multiple fusion states constitute an optimal state space.

10. Aircraft status monitoring system based on big data, characterized in that, including: a processor and a memory, where the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for monitoring the state of an aircraft based on big data according to any one of claims 1-9 is implemented.

Images