Method and system for automatic interpretation of aircraft telemetry data based on machine learning

Abstract

The application provides a kind of machine learning-based aircraft telemetry data automatic interpretation method and system, comprising: the telemetry data of multiple standard experiments is preprocessed;Based on the telemetry data after data preprocessing, template curve is constructed by improved K-means clustering method;By the method of segmented sequence matching and dynamic time adjustment, the experimental curve and template curve are compared, and the similarity of curve in each time period is calculated;The similarity of curve in each time period is compared with the set similarity threshold to realize the automatic interpretation of telemetry data.The technical scheme of the application is used to solve the technical problem that the existing technology cannot identify the trend anomaly or numerical floating anomaly of data in a certain time period.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of engineering applications and information science, and in particular to a method and system for automatic interpretation of aircraft telemetry data based on machine learning. Background Technology

[0002] In recent years, the number of aircraft model missions has increased year by year, and the testing tasks have become increasingly heavy, leading to a severe situation of human resource conflict and shortage. The data interpretation and analysis stage of the testing process is a crucial factor affecting the efficiency of test organization and the input of human resources. Improving the automation and intelligence of the test telemetry data interpretation and analysis stage, thereby improving test organization efficiency and reducing personnel input, has become a realistic and urgent need for all models.

[0003] Achieving intelligent analysis and interpretation of various test data has always been a key focus in the advancement of intelligent applications in aircraft model development. However, for many years, aerospace research institutes have failed to achieve a breakthrough. While various data integration and automatic interpretation systems are abundant and widely used in other industrial production sectors, their application is rarely seen in aircraft model development. The core reason lies in the difficulty of establishing a rule-based knowledge base for data interpretation and analysis. Conventional automatic data analysis and interpretation systems rely primarily on prior rule models and knowledge bases. However, aircraft development is highly customized; different models have different system compositions, workflows, and control laws, resulting in significant differences in the types, quantities, and characteristics of corresponding test data. Furthermore, during the development of most models, technical states and data parameters are adjusted. Each model requires significant investment of manpower and resources to establish and maintain its rule base, making the cost-effectiveness ratio low for a single model. Secondly, for some systems and specialties, the system's parameter characteristics and variation patterns often involve core intellectual property rights, which are not easily disclosed due to confidentiality and other factors. Therefore, constructing data interpretation and analysis rules using traditional technical approaches has always been a major obstacle to the automatic interpretation of aircraft telemetry data.

[0004] Currently, in practical applications, there are generally two methods to achieve automatic interpretation of telemetry data. The first is setting thresholds, which is the most common and widely used method. Its advantage lies in its simplicity and convenience. It only requires manual intervention to set an upper and lower limit for each parameter based on prior domain knowledge and simulation results, and then storing these limits in a database. After each simulation experiment, the experimental data is compared with the set upper and lower thresholds to analyze whether the data values ​​exceed the limits within a certain range. Users can also set alarm limits and abnormal limits according to actual conditions; this method is highly efficient.

[0005] Another approach is to calculate the envelope or parameter estimate. This involves using mathematical statistics to determine the envelope range of the curve, and then testing the experimental curve to determine if it fits the envelope. The advantage of this method is that it can update the envelope in real-time based on historical experimental data, resulting in higher accuracy compared to threshold methods.

[0006] While both methods can automate the interpretation of telemetry data to some extent, their primary function is to detect whether the data exceeds limits. They cannot identify abnormal trends or fluctuations in data over a given time period. Therefore, these methods can only assist researchers in interpreting data and cannot completely replace manual interpretation. Summary of the Invention

[0007] This invention provides a machine learning-based method and system for automatically interpreting aircraft telemetry data, which can solve the technical problem in the prior art of failing to identify abnormal trends or numerical fluctuations in data over a certain period of time.

[0008] According to one aspect of the present invention, an automatic interpretation method for aircraft telemetry data based on machine learning is provided. The automatic interpretation method for aircraft telemetry data based on machine learning includes: preprocessing telemetry data from multiple standard experiments; constructing a template curve based on the preprocessed telemetry data using an improved K-means clustering method; comparing the experimental curve and the template curve using segmented sequence matching and dynamic time adjustment methods to calculate the similarity of the curves at each time period; and comparing the similarity of the curves at each time period with a set similarity threshold to achieve automatic interpretation of the telemetry data.

[0009] Furthermore, based on the preprocessed telemetry data, the improved K-means clustering method is used to construct a template curve, specifically including: acquiring the preprocessed telemetry data; acquiring multiple initial cluster centers based on the preprocessed telemetry data; acquiring the distances between other points in the preprocessed telemetry data and each initial cluster center; assigning other points in the preprocessed telemetry data to the initial cluster centers closest to them; determining whether there are any telemetry data points in the experimental data that have not been assigned to a category; if all telemetry data points have been assigned, then recalculating the cluster centers of each cluster based on the data in the multiple clusters after the assignment, and determining whether the cluster centers of each cluster have changed; if the cluster center of any cluster has changed, then repeating the above process until the cluster centers of each cluster no longer change; and constructing a template curve based on the acquired cluster centers of each cluster.

[0010] Furthermore, a distance calculation formula based on dynamic time curvature is used to calculate the distances between the initial cluster centers and other points in the preprocessed telemetry data and each initial cluster center.

[0011] Furthermore, comparing the similarity of curves in each time period with a set similarity threshold to achieve automatic interpretation of telemetry data specifically includes: comparing the test curve with the template curve through segmented sequence matching and dynamic time adjustment; judging whether the test curve is abnormal through the similarity threshold index; if the similarity of a certain time period is lower than the set similarity threshold, an alarm is triggered and the abnormal curve segment and abnormal time point are output.

[0012] According to another aspect of the present invention, an automatic interpretation system for aircraft telemetry data is provided, which uses the machine learning-based automatic interpretation method for aircraft telemetry data described above to automatically interpret aircraft telemetry data.

[0013] Furthermore, the automatic interpretation system for aircraft telemetry data includes: a data preprocessing module, which preprocesses telemetry data from multiple standard experiments; a template curve construction module, which constructs template curves based on the preprocessed telemetry data using an improved K-means clustering method; a similarity calculation module, which compares the experimental curves and template curves using segmented sequence matching and dynamic time adjustment to calculate the similarity of the curves at each time interval; and an automatic interpretation module, which compares the similarity of the curves at each time interval with a set similarity threshold to achieve automatic interpretation of the telemetry data.

[0014] This invention provides an automatic interpretation method for aircraft telemetry data based on machine learning. This method analyzes telemetry data obtained from multiple repeated experiments, selects several relatively standard experiments as candidates, and constructs a template, also known as a standard curve, from the candidate experimental data using clustering methods. This is similar to constructing a standard "Chinese face" based on facial big data. When a new experimental curve needs to be detected, the experimental curve and the template curve can be compared using curve similarity matching to provide the similarity of the curves at various time periods. Users can set a similarity threshold according to their actual needs; exceeding the threshold will trigger an alarm, thus achieving automatic interpretation of telemetry data. DTW (Dynamic Time Warping) is incorporated into the curve similarity matching process to address the problems of unequal lengths and misalignment between experimental and template data, reducing false alarms caused by phase differences. To quickly identify abnormal time periods, a segmented matching weighted method is used, which not only evaluates the overall experiment but also quickly identifies abnormal time periods within the experiment. Therefore, compared with the prior art, the automatic interpretation method for aircraft telemetry data provided by this invention can effectively identify abnormal trends or abnormal numerical fluctuations in data within a certain period of time. By introducing machine learning technology, it effectively avoids the problem of establishing rule knowledge bases for each model separately. Using a general machine learning algorithm, it achieves a generalized data interpretation rule construction model based on existing data through learning and training. Attached Figure Description

[0015] The accompanying drawings, which form part of this specification, are provided to further illustrate embodiments of the invention and, together with the textual description, explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0016] Figure 1 A flowchart of an automatic interpretation method for aircraft telemetry data based on machine learning, according to a specific embodiment of the present invention, is shown.

[0017] Figure 2 A flowchart of the improved K-means clustering algorithm provided according to a specific embodiment of the present invention is shown. Detailed Implementation

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0020] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0021] like Figure 1 and Figure 2 As shown in the figure, a machine learning-based automatic interpretation method for aircraft telemetry data is provided according to a specific embodiment of the present invention. The machine learning-based automatic interpretation method for aircraft telemetry data includes: preprocessing telemetry data from multiple standard experiments; constructing a template curve based on the preprocessed telemetry data using an improved K-means clustering method; comparing the experimental curve and the template curve using segmented sequence matching and dynamic time adjustment methods to calculate the similarity of the curves at each time period; and comparing the similarity of the curves at each time period with a set similarity threshold to achieve automatic interpretation of the telemetry data.

[0022] This configuration method provides an automatic interpretation method for aircraft telemetry data based on machine learning. It analyzes telemetry data obtained from multiple repeated experiments, selects several relatively standard experiments as candidates, and constructs a template, also known as a standard curve, from the candidate experimental data using clustering methods. This is similar to constructing a standard "Chinese face" based on facial big data. When a new experimental curve needs to be detected, curve similarity matching is used to compare the experimental curve with the template curve, providing the similarity score at various time intervals. Users can set a similarity threshold according to their actual needs; exceeding the threshold triggers an alarm, thus achieving automatic interpretation of telemetry data. DTW (Dynamic Time Warping) is incorporated into the curve similarity matching process to address issues such as unequal lengths and misalignment between experimental and template data, reducing false alarms caused by phase differences. To quickly identify abnormal time intervals, a segmented matching weighted approach is used, which not only evaluates the overall experiment but also quickly identifies abnormal time intervals within the experiment. Therefore, compared with the prior art, the automatic interpretation method for aircraft telemetry data provided by this invention can effectively identify abnormal trends or abnormal numerical fluctuations in data within a certain period of time. By introducing machine learning technology, it effectively avoids the problem of establishing rule knowledge bases for each model separately. Using a general machine learning algorithm, it achieves a generalized data interpretation rule construction model based on existing data through learning and training.

[0023] Specifically, in this invention, to achieve automatic interpretation of aircraft telemetry data based on machine learning, it is first necessary to preprocess the telemetry data from multiple standard experiments. Data preprocessing mainly involves data alignment and outlier removal. Since the aircraft telemetry data is obtained through multiple flight experiments, it exhibits overall consistency. However, phase differences or inconsistencies in experiment duration may occur between the data curves from each experiment. Therefore, data alignment is necessary to find a unified starting point, facilitating the subsequent template creation.

[0024] Furthermore, after preprocessing the telemetry data from multiple standard experiments, a template curve can be constructed based on the preprocessed telemetry data using an improved K-means clustering method. In this invention, constructing the template curve based on the preprocessed telemetry data using the improved K-means clustering method specifically includes: acquiring the preprocessed telemetry data; acquiring multiple initial cluster centers based on the preprocessed telemetry data; acquiring the distances between other points in the preprocessed telemetry data and each initial cluster center; assigning the other points in the preprocessed telemetry data to the initial cluster center closest to them; determining whether any telemetry data points in the experimental data are not assigned to a category; if all telemetry data points have been assigned, recalculating the cluster centers of each cluster based on the data in the multiple clusters after the assignment, and determining whether the cluster centers of each cluster have changed; if the cluster center of any cluster changes, repeating the above process until the cluster centers of all clusters no longer change; and constructing the template curve based on the acquired cluster centers of each cluster.

[0025] As a specific embodiment of the present invention, at least five complete and anomaly-free experimental data sets are selected, and the improved K-means clustering algorithm is used to cluster each dimension of the experimental data separately. The working principle is as follows:

[0026] 1) By continuously iterating and dynamically adjusting cluster centers, data within the same cluster has high similarity, while data in different clusters has low similarity. Where: P represents the objective function, and K represents the number of clusters. d j Represents an element within this class, C i Indicates the cluster center of this class:

[0027]

[0028] 2) Improved the selection of initial cluster centers. When selecting a new cluster center, the K-means clustering algorithm makes it more likely that points farther away from existing cluster centers will be selected as cluster centers.

[0029]

[0030] Where X is the set of data in the class, x is a certain data in the class, C is the cluster center, and c is the cluster center of the class.

[0031] 3) Since the K-means algorithm is susceptible to noise and outliers due to the use of Euclidean distance, the distance calculation formula of the K-means clustering algorithm is modified to a distance calculation formula based on dynamic time warping (DTW).

[0032]

[0033] η(a,b)=X a -C b n +min

[0034] {η(a-1,b-1),η(a-1,b),η(a,b-1)}

[0035] Where X is the set of data in the class, C is the cluster center, and L is the cluster size. X Let L be the length of sequence X. C Let C be the length of sequence C. η(a,b) is the cumulative distance, ||·|| n For l n The norm, a and b, represents two points at different time points in the two time series that constitute the time series matrix. In this embodiment, a distance calculation formula based on dynamic time curvature is used to calculate the distances between the initial cluster centers and other points in the preprocessed telemetry data and each initial cluster center.

[0036] Furthermore, after constructing the template curve using the improved K-means clustering method, the experimental curve and the template curve can be compared using segmented sequence matching and dynamic time adjustment methods to calculate the similarity of the curves at each time period. The similarity of the curves at each time period is then compared with a set similarity threshold to achieve automatic interpretation of telemetry data.

[0037] In this invention, the automatic interpretation of telemetry data by comparing the similarity of curves in each time period with a set similarity threshold specifically includes: comparing the test curve with the template curve by segmented sequence matching and dynamic time adjustment, judging whether the test curve is abnormal by the similarity threshold index, and if the similarity of a certain time period is lower than the set similarity threshold, an alarm is triggered and the abnormal curve segment and abnormal time point are output.

[0038] This configuration method, by incorporating DTW (Dynamic Time Warping) into the curve similarity matching process, addresses the issues of unequal lengths and misalignment between experimental and template data, reducing false alarms caused by phase differences. To quickly identify anomalous time periods, a segmented matching weighting approach is used, enabling not only an overall evaluation of the experiment but also rapid identification of abnormal time periods. In this invention, segmented sequence matching weighting and dynamic time adjustment are commonly used methods in the prior art.

[0039] Specifically, in this invention, the main method for automatic data interpretation is dynamic time bending. This method matches the test curve and the template curve, and uses a similarity index to determine if the test curve has any anomalies. The core idea of ​​dynamic time bending is to bend the time axis, so that time-series data do not need to be matched one-to-one during the matching process, but only mapped. The similarity index is mainly considered in the sequence similarity judgment process. The specific process is as follows:

[0040] The high-dimensional time series generated by the spacecraft telemetry data are assumed to have two time series Q and C, with lengths n and m respectively. Here, Q is a pre-established template sequence, and C is a test sequence that needs to be matched.

[0041] Let Q = {q1, q2, ..., q} n}, C = {c1, c2, ..., c m The points in the two sequences form a time series matrix D. n*m = {d(i,j)}. The distance matrix D contains the distance between two points at different time points in the time series.

[0042] In matrix D, many paths can be found that satisfy the conditions. Suppose there is a path H = {h1, h2, ..., h...} t Let path H be the dynamic time-bending distance.

[0043]

[0044] The process of selecting the optimal path can be accomplished through dynamic programming. We can construct an accumulation matrix M = {m(i,j)} to record the shortest path from the starting position to the final position.

[0045]

[0046] According to another aspect of the present invention, an automatic interpretation system for aircraft telemetry data is provided, which uses the machine learning-based automatic interpretation method for aircraft telemetry data described above to automatically interpret aircraft telemetry data.

[0047] This configuration provides a machine learning-based automatic interpretation system for aircraft telemetry data. The system analyzes telemetry data from repeated experiments, selecting several relatively standard experiments as candidates. A template, also known as a standard curve, is constructed from these candidate experimental data using clustering methods—similar to creating a standard "Chinese face" based on facial big data. When a new experimental curve needs to be detected, curve similarity matching is used to compare the experimental curve with the template curve, providing the similarity score at various time intervals. Users can set similarity thresholds according to their needs; exceeding the threshold triggers an alarm, enabling automatic interpretation of telemetry data. DTW (Dynamic Time Warping) is incorporated into the curve similarity matching process to address issues of unequal lengths and misalignment between experimental and template data, reducing false alarms caused by phase differences. To quickly identify abnormal time intervals, a segmented matching weighted approach is used, enabling not only an overall evaluation of the experiment but also rapid identification of abnormal time intervals. Therefore, compared with the prior art, the automatic interpretation system for aircraft telemetry data provided by this invention can effectively identify abnormal trends or abnormal fluctuations in data within a certain period of time. By introducing machine learning technology, it effectively avoids the problem of building rule knowledge bases for each model separately. Using a general machine learning algorithm, it learns and trains on the basis of existing data to realize a general data interpretation rule construction model.

[0048] Specifically, in this invention, the automatic interpretation system for aircraft telemetry data includes a data preprocessing module, a template curve construction module, a similarity calculation module, and an automatic interpretation module. The data preprocessing module is used to preprocess telemetry data from multiple standard experiments. The template curve construction module is used to construct template curves based on the preprocessed telemetry data using an improved K-means clustering method. The similarity calculation module is used to compare the experimental curves and the template curves using segmented sequence matching and dynamic time adjustment methods to calculate the similarity of the curves at each time period. The automatic interpretation module is used to compare the similarity of the curves at each time period with a set similarity threshold to achieve automatic interpretation of the telemetry data.

[0049] To gain a further understanding of the present invention, the following description is provided in conjunction with... Figure 1 and Figure 2 The automatic interpretation method for aircraft telemetry data based on machine learning provided by this invention will be described in detail.

[0050] like Figure 1 and Figure 2As shown, this invention primarily addresses the practical needs of automatic interpretation and comprehensive analysis of telemetry data. It researches new methods and approaches for telemetry data processing using machine learning, aiming to effectively circumvent the challenge of establishing separate rule knowledge bases for each aircraft model by introducing machine learning technology. Utilizing general machine learning algorithms, based on existing data, a universal method for automatic interpretation of aircraft telemetry data is achieved through learning and training. This method utilizes telemetry data from aircraft subsystems to detect anomalies in different systems, representing a multi-parameter anomaly detection method for subsystems. This method can identify anomalies that are difficult to detect using traditional methods, achieving multi-parameter anomaly detection for spacecraft subsystems, and has good prospects for widespread application.

[0051] The technical solution of this invention, the method for automatically interpreting telemetry data and constructing templates, mainly consists of three parts. The first part is the data preprocessing module, which primarily performs data alignment and outlier removal on the telemetry data. Aircraft telemetry data is obtained through multiple flight experiments, so it exhibits a consistent overall trend. However, phase differences or inconsistent experiment durations may occur between the data curves from each experiment. Therefore, data alignment is necessary to find a unified starting point for subsequent template construction. The second part is the template construction module, where the preprocessed data is input for template construction. This part requires data from at least five complete, repeated experiments, ensuring it is free of anomalies and that the number of parameters is consistent across experiments. Subsequently, clustering is performed using the K-means algorithm. This invention improves upon the traditional K-means algorithm, enhancing its efficiency. After clustering, a standard curve is obtained and stored in the database as a template. During subsequent data interpretation, the test curve is matched with the template curve for similarity. The third part is the automatic telemetry data interpretation module. The system takes a test curve as input and compares it with a template curve using segmented sequence matching. Based on a predefined similarity threshold, the system returns the detection results to the user. If the similarity falls below the threshold for a given time period, the system will issue an alarm and output the abnormal curve segment and the abnormal time point.

[0052] The advantages of this invention compared to existing technologies are as follows: Addressing the current problems in the automatic interpretation of aircraft telemetry data, this invention proposes a template-based automatic interpretation method. This method preprocesses the raw data, uses an improved K-means clustering algorithm to find cluster centers, constructs templates for all system parameters, and stores them in a database. Test parameters are compared with template curves using segmented matching and dynamic time warping, and anomalies are determined based on similarity and other metrics. In practical applications, the template-based method allows for flexible parameter selection, enabling simultaneous testing of multiple parameters and cross-system parameter selection. It is more flexible and practical than existing methods and can be extended to other systems.

[0053] This invention utilizes telemetry data from aircraft for anomaly detection. The data is preprocessed, and then a modified K-means clustering algorithm is used to construct a template from the input data, which is stored in a database. Test data is then input into an automatic interpretation system. Dynamic time warping is used to obtain the similarity between the test data and the template data. Based on a set threshold, a system anomaly is determined, thereby achieving the goal of detecting aircraft system anomalies. The specific implementation steps are as follows:

[0054] Step 1: Perform data preprocessing on the telemetry data from multiple standard experiments.

[0055] Step two: Based on the preprocessed telemetry data, a template curve is constructed using an improved K-means clustering method. Specifically, in this embodiment, preprocessed telemetry data is acquired; multiple initial cluster centers are obtained based on the preprocessed telemetry data; the distances between other points in the preprocessed telemetry data and each initial cluster center are obtained; the other points in the preprocessed telemetry data are assigned to the initial cluster centers closest to them; it is determined whether there are any telemetry data points in the experimental data that have not been assigned to a category. If all telemetry data points have been assigned, the cluster centers of each cluster are recalculated based on the data in the multiple clusters after the assignment is completed, and it is determined whether the cluster centers of each cluster have changed. If the cluster center of any cluster changes, the above process is repeated until the cluster centers of each cluster no longer change; a template curve is constructed based on the obtained cluster centers of each cluster.

[0056] Step three involves comparing the experimental curve and the template curve using segmented sequence matching and dynamic time adjustment to calculate the similarity of the curves at each time interval. In this embodiment, the main method for automatic data interpretation is dynamic time bending. This method is used to match the test curve and the template curve, and the similarity index is used to determine whether the test curve has any anomalies. The core idea of ​​dynamic time bending is to bend the time axis so that the time series data do not need to be matched one-to-one during the matching process, but only mapped. The similarity index is mainly considered in the sequence similarity judgment process.

[0057] Step four involves comparing the similarity of the curves across different time periods with a set similarity threshold to achieve automatic interpretation of telemetry data. In this embodiment, the test curve is compared with the template curve using segmented sequence matching and dynamic time adjustment. The similarity threshold is used to determine if the test curve is abnormal. If the similarity of a certain time period is lower than the set similarity threshold, an alarm is triggered and the abnormal curve segment and abnormal time point are output.

[0058] In summary, this invention provides an automatic interpretation method for aircraft telemetry data based on machine learning. This method analyzes telemetry data obtained from multiple repeated experiments, selects several relatively standard experiments as candidates, and constructs a template, also known as a standard curve, from the candidate experimental data using clustering methods. This is similar to constructing a standard "Chinese face" based on facial big data. When a new experimental curve needs to be detected, the experimental curve and the template curve can be compared using curve similarity matching to provide the similarity of the curves at various time periods. Users can set a similarity threshold according to their actual needs; exceeding the similarity threshold will trigger an alarm, thus achieving automatic interpretation of telemetry data. DTW (Dynamic Time Warping) is incorporated into the curve similarity matching process to address the problems of unequal lengths and misalignment between experimental and template data, reducing false alarms caused by phase differences. To quickly identify abnormal time periods, a segmented matching weighted approach is used, which not only evaluates the overall experiment but also quickly identifies abnormal time periods within the experiment. Therefore, compared with the prior art, the automatic interpretation method for aircraft telemetry data provided by this invention can effectively identify abnormal trends or abnormal numerical fluctuations in data within a certain period of time. It applies artificial intelligence technologies such as machine learning to the processing and analysis of experimental data parameters, extracts data features from the data itself, and replaces the traditional manually set rules, fundamentally solving the problem that the rules for automatic analysis and comprehensive processing of telemetry data cannot be accurately described and it is difficult to build deterministic models.

[0059] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.

[0060] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore should not be construed as limiting the scope of protection of this invention.

[0061] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for automatic interpretation of aircraft telemetry data based on machine learning, characterized in that, The automatic interpretation method for aircraft telemetry data based on machine learning includes: Data preprocessing was performed on telemetry data from multiple standard experiments; Based on the telemetry data after data preprocessing, a template curve is constructed using an improved K-means clustering method. The improved K-means clustering method improves the selection of the initial center point of the algorithm, so that when the K-means clustering algorithm selects a new cluster center, the point that is farther away from the existing cluster center has a greater probability of being selected as the cluster center. The experimental curve and the template curve are compared by segmented sequence matching and dynamic time adjustment, and the similarity of the curves at each time period is calculated. The similarity of curves in different time periods is compared with a set similarity threshold to achieve automatic interpretation of telemetry data.

2. The automatic interpretation method for aircraft telemetry data based on machine learning according to claim 1, characterized in that, Based on the preprocessed telemetry data, a template curve is constructed using an improved K-means clustering method, specifically including: Acquire preprocessed telemetry data; Multiple initial cluster centers were obtained based on the preprocessed telemetry data; Obtain the distances between other points in the preprocessed telemetry data and each initial cluster center; Other points in the preprocessed telemetry data are assigned to the initial cluster centers that are closest to them; Determine if there are any telemetry data points in the experimental data that have not been assigned to a category. If all telemetry data points have been assigned to a category, then recalculate the cluster center of each cluster based on the data in the multiple clusters after the division, and determine whether the cluster center of each cluster has changed. If the cluster center of any cluster changes, repeat the above process until the cluster center of each cluster no longer changes. Template curves are constructed based on the cluster centers of each obtained cluster.

3. The automatic interpretation method for aircraft telemetry data based on machine learning according to claim 2, characterized in that, The distance between other points in the preprocessed telemetry data and each initial cluster center is calculated using a distance calculation formula based on dynamic time curvature.

4. The automatic interpretation method for aircraft telemetry data based on machine learning according to claim 3, characterized in that, The automatic interpretation of telemetry data is achieved by comparing the similarity of curves at different time periods with a set similarity threshold. Specifically, this includes comparing the experimental curve with the template curve through segmented sequence matching and dynamic time adjustment, judging whether there are any abnormalities in the experimental curve through the similarity threshold index, and if the similarity of a certain time period is lower than the set similarity threshold, an alarm is triggered and the abnormal curve segment and abnormal time point are output.

5. An automatic interpretation system for aircraft telemetry data, characterized in that, The automatic interpretation system for aircraft telemetry data uses the machine learning-based automatic interpretation method for aircraft telemetry data as described in any one of claims 1 to 4 to perform automatic interpretation of aircraft telemetry data.

6. The automatic interpretation system for aircraft telemetry data according to claim 5, characterized in that, The automatic interpretation system for aircraft telemetry data includes: A data preprocessing module is used to preprocess telemetry data from multiple standard experiments. The template curve construction module is used to construct template curves based on telemetry data after data preprocessing using an improved K-means clustering method. The improved K-means clustering method improves the selection of the initial center point of the algorithm, so that when the K-means clustering algorithm selects a new cluster center, the point farther away from the existing cluster center has a higher probability of being selected as the cluster center. The similarity calculation module is used to compare the experimental curve and the template curve by segmented sequence matching and dynamic time adjustment, and to calculate the similarity of the curves at each time period. An automatic interpretation module is used to compare the similarity of curves in each time period with a set similarity threshold to achieve automatic interpretation of telemetry data.

Images