Aviation big data intelligent analysis method based on trajectory anomaly detection

Abstract

The invention relates to an aviation big data intelligent analysis method based on trajectory anomaly detection, and belongs to the technical field of aviation data processing. The method comprises the following steps: acquiring operation state data and flight parameters of an aircraft, and extracting aerodynamic parameters of an aircraft type and deviation data corresponding to horizontal navigation and vertical navigation; carrying out stream batch integrated processing on the data through a real-time processing engine, and carrying out space-time reference unification to obtain standardized data; performing data cleaning, abnormal point correction and missing value interpolation on the standardized data; performing pneumatic-performance correlation analysis after data processing, and quantifying the influence of parameter deviation on the climbing performance through a dynamic response model to obtain a strong correlation coefficient; performing coupling analysis through the climbing rate attenuation prediction model to obtain an aircraft climbing performance anomaly prediction result; and generating a correction instruction according to the persistent climb rate anomaly of the performance anomaly prediction result. The efficient and accurate analysis of the aviation big data is realized, and the safety and reliability of the operation of the aircraft are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of aviation data processing, and in particular relates to an intelligent analysis method of aviation big data based on trajectory anomaly detection. Background Art

[0002] The application of trajectory information-based anomaly detection technology in the aviation field mainly relies on the collection and processing of aircraft operation data (such as ACARS messages and QAR parameters), combined with aerodynamic performance analysis, spatiotemporal data modeling and machine learning algorithms to achieve early warning of aircraft climb performance anomalies.

[0003] There are still areas for improvement:

[0004] The aircraft transmits operational status data in real time through the ACARS system, while the QAR recorder stores high-precision flight parameters (such as airspeed, thrust, navigation deviation, etc.). Traditional methods rely on batch processing frameworks for data cleaning and fusion, but face the problem of inconsistent spatiotemporal benchmarks between streaming data and batch data, resulting in insufficient real-time performance; the timestamps and coordinate systems of ACARS streaming data and QAR batch data are inconsistent (such as GNSS and inertial navigation data deviation), resulting in broken or jumpy trajectory sequences, affecting the accuracy of anomaly detection;

[0005] Based on aerodynamic differential equations and state-space models, the impact of flight parameter deviations on climb performance is quantified. Existing methods mostly use static aerodynamic envelope analysis, which is difficult to dynamically adapt to changes in aerodynamic characteristics in different flight phases (such as climb and cruise), and lack multi-source coupling analysis with meteorological and route structure.

[0006] Traditional anomaly detection relies on statistical thresholds or clustering algorithms (such as DBSCAN). However, due to the high dimensionality, temporal nature, and multi-source heterogeneity of aviation data (such as the coordinate system differences between inertial navigation data and GNSS data), detection accuracy and real-time performance are limited.

[0007] After an abnormal warning, graded correction instructions need to be generated (such as adjusting thrust and reconstructing the navigation profile). Existing systems rely heavily on manual experience and lack automated closed-loop control capabilities, resulting in delayed responses. Summary of the Invention

[0008] In order to solve the above problems existing in the prior art, the present invention provides an intelligent analysis method of aviation big data based on trajectory anomaly detection.

[0009] The purpose of the present invention can be achieved through the following technical solutions:

[0010] S1: Acquire aircraft operating status data through the aircraft communication addressing and reporting system; obtain flight parameters from the quick access recorder, extract aircraft model aerodynamic parameters and corresponding horizontal navigation and vertical navigation deviation data from the navigation database in the flight management system; use the real-time processing engine to process the aircraft operating status data, flight parameters, aircraft model aerodynamic parameters, and deviation data in a batch-processing manner and unify them in time and space to obtain standardized data;

[0011] S2: performing data cleaning on the standardized data, identifying and correcting abnormal points in the standardized data by combining the aircraft aerodynamic characteristics and the distribution of historical flight data, and interpolating missing values ​​for flight data signal loss in the standardized data;

[0012] S3: Perform an aerodynamic-performance correlation analysis based on the cleaned data. Quantify the impact of parameter deviations on climb performance using a dynamic response model to obtain a strong correlation coefficient between key flight parameters and climb altitude anomalies. A climb rate attenuation prediction model is then used to couple this strong correlation coefficient with meteorological data and route structure data to obtain an aircraft climb performance anomaly prediction result.

[0013] S4: Generate a correction instruction according to the sustained climb rate anomaly in the performance anomaly prediction result.

[0014] Specifically, the method for integrated flow and batch processing is:

[0015] S101: Build a hybrid computing architecture that accesses streaming data from the aircraft communication addressing and reporting system in real time through a distributed message queue, while also loading batch data from the flight data in the fast access recorder by establishing a micro-batch processing channel;

[0016] S102: Align the time axis of stream and batch data based on event time semantics, apply a sliding time window function to the streaming data of the aircraft communication addressing and reporting system to achieve aggregation, perform segmented resampling based on flight phase identifiers on the batch data of the fast access recorder, and complete the space-time reference unification using the WGS-84 coordinate system and UTC timestamp;

[0017] S103: Maintain the real-time dynamic context information of the aircraft in the Flink state management mechanism. The complex event processing engine performs pattern recognition on flight profile switching events. When the start point of the climb phase is detected, the batch processing layer is triggered to call the pre-stored aircraft aerodynamic parameter library and perform dynamic correlation calculations with the real-time streaming data.

[0018] S104: Implement unified registration and management of multi-source data formats through the Model Registration Center. Use protocol buffers to perform structured deserialization of aircraft communication addressing and reporting system messages. Implement binary parsing based on the civil aviation standard flight data frame structure for fast access recorder parameter sets. Ultimately, construct a four-dimensional feature standardized data cube containing timestamps, three-dimensional spatial coordinates, flight phase status, and dynamic operating parameters.

[0019] Specifically, the method for unifying the time and space references includes:

[0020] The embedded Global Positioning System (GPS) timestamps are extracted from Aircraft Communications Addressing and Reporting System (ACARS) streaming data, and the absolute flight time recorded in the Quick Access Recorder (QAR) batch data is parsed. The two time sources are calibrated using the Network Time Protocol (NTP) and converted to a unified Coordinated Universal Time (UTC) standard time series. The NTP protocol uses a hierarchical synchronization architecture to correct device clock deviations with millisecond-level accuracy, ensuring time consistency across systems. An event-time-based dual-streamline mechanism is established to dynamically compensate for transmission delays between streaming real-time data and batch historical data. Time window correlation technology is used to align real-time data streams with historical batch data along the time dimension.

[0021] Spatial datum conversion: Convert the longitude and latitude coordinates in the ACARS position report from the WGS-84 coordinate system to the Earth-centered, Earth-fixed, three-dimensional rectangular coordinate system. Normalize the coordinate system of the inertial navigation system data recorded by the QAR. Use the coordinate rotation matrix to eliminate the datum deviation between different navigation sources and generate a unified spatial trajectory sequence.

[0022] Spatiotemporal interpolation alignment: To address the sampling discrepancies between high-frequency stream data and low-frequency batch data, the Akima spline interpolation algorithm is used to resample the QAR parameters in the time dimension under a unified spatiotemporal benchmark. In combination with the flight phase identifier, piecewise linear interpolation based on the flight segment profile is implemented in the spatial dimension to achieve sub-second alignment of stream and batch data within a four-dimensional spatiotemporal grid.

[0023] Dynamic benchmarking: This system fuses multi-source position data through a Kalman filter to construct a multi-dimensional state vector (longitude, latitude, barometric altitude, and GNSS altitude). It then predicts trajectory continuity based on the aircraft's kinematic model. When a coordinate system jump or timestamp breakpoint is detected, dead reckoning based on heading angle and ground speed is triggered for data compensation.

[0024] Spatiotemporal associative storage: The unified spatiotemporal benchmark data is encoded into a composite key structure (ICAO address code, UTC whole second time, geographic grid block). A three-level spatiotemporal index is established in the distributed storage system. The geographic grid blocks use the H3 discrete global grid system for hexagonal spatial sharding, supporting spatiotemporal range queries across stream and batch datasets.

[0025] Specifically, the method for identifying and correcting abnormal points in the standardized data based on the distribution of aircraft aerodynamic characteristics and flight segment historical data is as follows:

[0026] S201: Based on the lift coefficient, drag coefficient, and engine thrust curves in the aircraft aerodynamic parameter library, combined with the current flight altitude and Mach number, the aerodynamic performance envelope boundaries are dynamically generated. Envelope analysis is performed on the angle of attack, airspeed, and thrust-to-weight ratio parameters in the standardized data. If a point exceeds the aerodynamic performance envelope threshold for three consecutive sampling periods, it is marked as an aerodynamic characteristic abnormality point.

[0027] S202: Filter historical flight data of the same route and the same aircraft configuration from the flight segment history database to construct a benchmark data set; achieve nonlinear alignment of the flight phase time axis through a dynamic time warping algorithm to eliminate differences in flight duration, phase transition rhythm, etc. between different flight segments; use a density-based ordered clustering algorithm to perform feature analysis on multidimensional flight parameters (including vertical speed, heading angle, etc.) to construct multidimensional distribution characteristics of the flight segment parameters; calculate the Mahalanobis distance of the current standardized parameter based on the covariance matrix of the cluster feature space. When the observed value deviates from the Mahalanobis distance of the cluster center by more than 3 standard deviations, a flight segment distribution abnormality warning is triggered.

[0028] S203: An extended Kalman filter is used to perform state estimation on aerodynamic characteristic anomalies. Redundant measurements from the inertial navigation system and the atmospheric data system are integrated to generate a residual sequence. When the Frobenius norm of the residual covariance matrix exceeds a preset threshold, a reverse deduction correction based on the aerodynamic equation is triggered. For flight segment distribution anomalies, a multi-source verification channel is constructed by correlating weather radar data with Automatic Dependent Surveillance-Broadcast (ADS-B) system surveillance data. If external data sources verify that the anomaly persists, the original data is retained and a credibility label is added.

[0029] Specifically, the missing value interpolation method is:

[0030] Based on the flight phase identifier and spatiotemporal reference data, the flight phase in which the rapid access recorder signal is lost is identified, and a differentiated interpolation strategy is selected according to the phase characteristics.

[0031] For data windows where the duration of continuous missing data is shorter than the duration of stability, interpolation is performed based on adjacent valid data points, and the smoothing coefficient is dynamically calculated through the sliding window to adapt to the data fluctuation characteristics;

[0032] For data windows whose continuous missing duration exceeds the stability duration and are in the stable cruising stage, a pattern matching method based on historical flight data is used to retrieve historical data segments that match the flight stages before and after the current missing data window from the flight segment history database, extract the time series of corresponding parameters in the historical data segments, and match the historical data segments that are most similar to the missing data window through the time series similarity measurement algorithm. Based on this, linear interpolation or spline interpolation method is used to interpolate missing values.

[0033] Specifically, the aerodynamic-performance correlation analysis method is:

[0034] S301: Construct an aerodynamic differential equation model based on the aircraft model aerodynamic parameter library, associate the lift coefficient, drag coefficient with flight state parameters, and establish a dynamic response relationship:

[0035]

[0036] Where T is the engine thrust, D is the aerodynamic drag, L is the lift, W is the weight of the aircraft, γ is the climb angle; α is the angle of attack, Ma is the Mach number, and H is the altitude;

[0037] The partial derivatives of aerodynamic parameters on climb rate are calculated through parameter sensitivity analysis to quantify the influence of each parameter deviation on performance.

[0038] S302: Using a state-space model to characterize the longitudinal motion of the aircraft, using QAR parameter deviations (e.g., N1 rotation speed deviation ΔN1, airspeed deviation ΔV) as input disturbances and climb altitude deviation ΔH as output responses;

[0039] S303: Design a sliding time test to dynamically identify the causal relationship strength between parameter deviation and performance anomaly, and screen out the key QAR parameter set with a strong correlation coefficient (|r|>0.8).

[0040] Specifically, the dynamic response model integrates the engine performance attenuation factor with the wing surface pressure distribution data, and calculates the aerodynamic efficiency loss through finite element analysis; introduces meteorological data as boundary conditions, and uses computational fluid dynamics transient simulation to evaluate the disturbance effect of the external environment on the climbing performance.

[0041] Specifically, the calculation method of the climb rate attenuation prediction model is:

[0042] A spatiotemporal graph convolutional network is established. It uses the four-dimensional features of a standardized data cube as input to construct a graph structure containing aerodynamic topological relationships. The nodes represent flight state parameters, and the edge weights are determined by the aerodynamic coupling coefficient.

[0043] Train a multi-task learning model to simultaneously predict the climb rate attenuation trend and anomaly level. The loss function is designed as:

[0044]

[0045] Among them, λ1, λ2 are task weight coefficients, MSE is mean square error loss, CrossEntropy is cross entropy loss, y RoC is the target climb rate value of the sample, Represents the target climb rate value predicted by the model, y class is the actual abnormality level label, represents the probability distribution of anomaly levels predicted by the model;

[0046] An online incremental learning engine is deployed to dynamically update the aerodynamic-performance correlation matrix when new flight segment data is detected, and random forest feature importance ranking is used to correct strong correlation coefficients in real time.

[0047] Specifically, the sustained climb rate anomaly is an aircraft performance anomaly judgment model based on multi-dimensional feature fusion. By integrating multi-source information such as aerodynamic performance, meteorological conditions, and route structure, it provides early warning of the performance degradation trend during the aircraft's climb process.

[0048] Specifically, the correction instructions include primary instructions, intermediate instructions, and advanced instructions; the primary instructions are to adjust the N1 limit value of the thrust management system, the intermediate instructions are to reconstruct the FMS vertical navigation profile, and the advanced instructions are to trigger a maintenance warning.

[0049] The beneficial effects of the present invention are:

[0050] This system achieves sub-second spatiotemporal alignment of ACARS streaming data and QAR batch data through a stream-batch integrated processing architecture, thereby improving the speed of QAR parsing. Based on aerodynamic performance envelope boundary detection and segment distribution Mahalanobis distance analysis, the accuracy of outlier identification is improved. Combined with Holt double exponential smoothing and low-rank matrix decomposition technology, the compliance rate of missing value interpolation is improved. By constructing an aerodynamic-state space coupling model to quantify the impact of parameter deviations, a mixed integer programming model that integrates meteorological and route constraints is used to predict climb anomalies. When the rate of climb is continuously detected, multi-dimensional correction instructions including thrust compensation, attitude adjustment, and route optimization can be quickly generated, providing real-time, accurate, and intelligent decision-making support for aircraft performance monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0052] Figure 1 Schematic diagram of the process of the aviation big data intelligent analysis method based on trajectory anomaly detection of the present invention;

[0053] Figure 2This is a structural block diagram of the climb rate attenuation prediction model in the present invention. DETAILED DESCRIPTION

[0054] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0055] See also Figure 1 , an intelligent analysis method of aviation big data based on trajectory anomaly detection, including:

[0056] S1: Acquire aircraft operating status data through the aircraft communication addressing and reporting system; obtain flight parameters from the quick access recorder, extract aircraft model aerodynamic parameters and corresponding horizontal navigation and vertical navigation deviation data from the navigation database in the flight management system; use the real-time processing engine to process the aircraft operating status data, flight parameters, aircraft model aerodynamic parameters, and deviation data in a batch-processing manner and unify them in time and space to obtain standardized data;

[0057] S2: performing data cleaning on the standardized data, identifying and correcting abnormal points in the standardized data by combining the aircraft aerodynamic characteristics and the distribution of historical flight data, and interpolating missing values ​​for flight data signal loss in the standardized data;

[0058] S3: Perform an aerodynamic-performance correlation analysis based on the cleaned data. Quantify the impact of parameter deviations on climb performance using a dynamic response model to obtain a strong correlation coefficient between key flight parameters and climb altitude anomalies. A climb rate attenuation prediction model is then used to couple this strong correlation coefficient with meteorological data and route structure data to obtain an aircraft climb performance anomaly prediction result.

[0059] S4: Generate a correction instruction according to the sustained climb rate anomaly in the performance anomaly prediction result.

[0060] Specifically, the method for integrated flow and batch processing is:

[0061] S101: Build a hybrid computing architecture that accesses streaming data from the aircraft communication addressing and reporting system in real time through a distributed message queue, while also loading batch data from the flight data in the fast access recorder by establishing a micro-batch processing channel;

[0062] S102: Align the time axis of stream and batch data based on event time semantics, apply a sliding time window function to the streaming data of the aircraft communication addressing and reporting system to achieve aggregation, perform segmented resampling based on flight phase identifiers on the batch data of the fast access recorder, and complete the space-time reference unification using the WGS-84 coordinate system and UTC timestamp;

[0063] S103: Maintain the real-time dynamic context information of the aircraft in the Flink state management mechanism. The complex event processing engine performs pattern recognition on flight profile switching events. When the start point of the climb phase is detected, the batch processing layer is triggered to call the pre-stored aircraft aerodynamic parameter library and perform dynamic correlation calculations with the real-time streaming data.

[0064] S104: Implement unified registration and management of multi-source data formats through the Model Registration Center. Use protocol buffers to perform structured deserialization of aircraft communication addressing and reporting system messages. Implement binary parsing based on the civil aviation standard flight data frame structure for fast access recorder parameter sets. Ultimately, construct a four-dimensional feature standardized data cube containing timestamps, three-dimensional spatial coordinates, flight phase status, and dynamic operating parameters.

[0065] In this embodiment, a computing engine that supports both stream and batch processing modes is used to uniformly access the aircraft operating status data transmitted in real time by ACARS and the flight parameters, aerodynamic parameters, and navigation deviation data recorded by QAR. The aircraft operating status data is stream data, and the flight parameters, aerodynamic parameters, and navigation deviation data are batch data. Stream data is ingested in real time through message queues such as Kafka, and batch data is loaded from HDFS or a distributed database. The two types of data are processed in parallel in a unified computing engine. Stream data is divided into sliding windows (e.g., 5-second windows) based on event time to dynamically process real-time data streams. Batch data is divided into fixed windows based on flight segment periods (e.g., a single flight mission) to perform offline batch calculations. The intermediate states across windows are managed through the state backend (RocksDB) to ensure state consistency of stream and batch tasks.

[0066] Specifically, the method for unifying the time and space references includes:

[0067] The embedded Global Positioning System (GPS) timestamps are extracted from Aircraft Communications Addressing and Reporting System (ACARS) streaming data, and the absolute flight time recorded in the Quick Access Recorder (QAR) batch data is parsed. The two time sources are calibrated using the Network Time Protocol (NTP) and converted to a unified Coordinated Universal Time (UTC) standard time series. The NTP protocol uses a hierarchical synchronization architecture to correct device clock deviations with millisecond-level accuracy, ensuring time consistency across systems. An event-time-based dual-streamline mechanism is established to dynamically compensate for transmission delays between streaming real-time data and batch historical data. Time window correlation technology is used to align real-time data streams with historical batch data along the time dimension.

[0068] Spatial datum conversion: Convert the longitude and latitude coordinates in the ACARS position report from the WGS-84 coordinate system to the Earth-centered, Earth-fixed, three-dimensional rectangular coordinate system. Normalize the coordinate system of the inertial navigation system data recorded by the QAR. Use the coordinate rotation matrix to eliminate the datum deviation between different navigation sources and generate a unified spatial trajectory sequence.

[0069] Spatiotemporal interpolation alignment: To address the sampling discrepancies between high-frequency stream data and low-frequency batch data, the Akima spline interpolation algorithm is used to resample the QAR parameters in the time dimension under a unified spatiotemporal benchmark. In combination with the flight phase identifier, piecewise linear interpolation based on the flight segment profile is implemented in the spatial dimension to achieve sub-second alignment of stream and batch data within a four-dimensional spatiotemporal grid.

[0070] Dynamic benchmarking: This system fuses multi-source position data through a Kalman filter to construct a multi-dimensional state vector (longitude, latitude, barometric altitude, and GNSS altitude). It then predicts trajectory continuity based on the aircraft's kinematic model. When a coordinate system jump or timestamp breakpoint is detected, dead reckoning based on heading angle and ground speed is triggered for data compensation.

[0071] Spatiotemporal associative storage: The unified spatiotemporal benchmark data is encoded into a composite key structure (ICAO address code, UTC whole second time, geographic grid block). A three-level spatiotemporal index is established in the distributed storage system. The geographic grid blocks use the H3 discrete global grid system for hexagonal spatial sharding, supporting spatiotemporal range queries across stream and batch datasets.

[0072] Specifically, the method for identifying and correcting abnormal points in the standardized data based on the distribution of aircraft aerodynamic characteristics and flight segment historical data is as follows:

[0073] S201: Based on the lift coefficient, drag coefficient, and engine thrust curves in the aircraft aerodynamic parameter library, combined with the current flight altitude and Mach number, the aerodynamic performance envelope boundaries are dynamically generated. Envelope analysis is performed on the angle of attack, airspeed, and thrust-to-weight ratio parameters in the standardized data. If a point exceeds the aerodynamic performance envelope threshold for three consecutive sampling periods, it is marked as an aerodynamic characteristic abnormality point.

[0074] S202: Filter historical flight data of the same route and the same aircraft configuration from the flight segment history database to construct a benchmark data set; achieve nonlinear alignment of the flight phase time axis through a dynamic time warping algorithm to eliminate differences in flight duration, phase transition rhythm, etc. between different flight segments; use a density-based ordered clustering algorithm to perform feature analysis on multidimensional flight parameters (including vertical speed, heading angle, etc.) to construct multidimensional distribution characteristics of the flight segment parameters; calculate the Mahalanobis distance of the current standardized parameter based on the covariance matrix of the cluster feature space. When the observed value deviates from the Mahalanobis distance of the cluster center by more than 3 standard deviations, a flight segment distribution abnormality warning is triggered.

[0075] S203: An extended Kalman filter is used to perform state estimation on aerodynamic characteristic anomalies. Redundant measurements from the inertial navigation system and the atmospheric data system are integrated to generate a residual sequence. When the Frobenius norm of the residual covariance matrix exceeds a preset threshold, a reverse deduction correction based on the aerodynamic equation is triggered. For flight segment distribution anomalies, a multi-source verification channel is constructed by correlating weather radar data with Automatic Dependent Surveillance-Broadcast (ADS-B) system surveillance data. If external data sources verify that the anomaly persists, the original data is retained and a credibility label is added.

[0076] Specifically, the missing value interpolation method is:

[0077] Based on the flight phase identifier and spatiotemporal reference data, the flight phase in which the rapid access recorder signal is lost is identified, and a differentiated interpolation strategy is selected according to the phase characteristics.

[0078] For data windows where the duration of continuous missing data is shorter than the duration of stability, interpolation is performed based on adjacent valid data points, and the smoothing coefficient is dynamically calculated through the sliding window to adapt to the data fluctuation characteristics;

[0079] For data windows whose continuous missing duration exceeds the stability duration and are in the stable cruising stage, a pattern matching method based on historical flight data is used to retrieve historical data segments that match the flight stages before and after the current missing data window from the flight segment history database, extract the time series of corresponding parameters in the historical data segments, and match the historical data segments that are most similar to the missing data window through the time series similarity measurement algorithm. Based on this, linear interpolation or spline interpolation method is used to interpolate missing values.

[0080] Specifically, the aerodynamic-performance correlation analysis method is:

[0081] S301: Construct an aerodynamic differential equation model based on the aircraft model aerodynamic parameter library, associate the lift coefficient, drag coefficient with flight state parameters, and establish a dynamic response relationship:

[0082]

[0083] Where T is the engine thrust, D is the aerodynamic drag, L is the lift, W is the weight of the aircraft, γ is the climb angle; α is the angle of attack, Ma is the Mach number, and H is the altitude;

[0084] The partial derivatives of aerodynamic parameters on climb rate are calculated through parameter sensitivity analysis to quantify the influence of each parameter deviation on performance.

[0085] S302: Using a state-space model to characterize the longitudinal motion of the aircraft, using QAR parameter deviations (e.g., N1 rotation speed deviation ΔN1, airspeed deviation ΔV) as input disturbances and climb altitude deviation ΔH as output responses;

[0086] S303: Design a sliding time test to dynamically identify the causal relationship strength between parameter deviation and performance anomaly, and screen out the key QAR parameter set with a strong correlation coefficient (|r|>0.8).

[0087] Specifically, the dynamic response model integrates the engine performance attenuation factor with the wing surface pressure distribution data, and calculates the aerodynamic efficiency loss through finite element analysis; introduces meteorological data as boundary conditions, and uses computational fluid dynamics transient simulation to evaluate the disturbance effect of the external environment on the climbing performance.

[0088] Specifically, the calculation method of the climb rate attenuation prediction model is:

[0089] A spatiotemporal graph convolutional network is established. It uses the four-dimensional features of a standardized data cube as input to construct a graph structure containing aerodynamic topological relationships. The nodes represent flight state parameters, and the edge weights are determined by the aerodynamic coupling coefficient.

[0090] Train a multi-task learning model to simultaneously predict the climb rate attenuation trend and anomaly level. The loss function is designed as:

[0091]

[0092] Among them, λ1, λ2 are task weight coefficients, MSE is mean square error loss, CrossEntropy is cross entropy loss, y RoC is the target climb rate value of the sample, Represents the target climb rate value predicted by the model, y class is the actual abnormality level label, represents the probability distribution of anomaly levels predicted by the model;

[0093] An online incremental learning engine is deployed to dynamically update the aerodynamic-performance correlation matrix when new flight segment data is detected, and random forest feature importance ranking is used to correct strong correlation coefficients in real time.

[0094] In this embodiment, Figure 2 As shown, based on GRIB meteorological analysis and segment interpolation, the route gradient limit (climb angle threshold) is calculated, and a route structure diagram is constructed as the input of the graph neural network. The equivalent drag increment is calculated by decomposing the wind speed / flow direction into parallel / perpendicular components. Based on S102 sliding window aggregation (ACARS flow data) and QAR segment resampling, the UTC timestamp is unified. Discrete meteorological points are mapped to the track grid through quadratic Lagrangian interpolation. Based on the multi-task learning network, a coupled feature pool is constructed: the input layer is aerodynamic parameter deviation, wind disturbance, and route gradient constraint. The attention mechanism of the gate control unit dynamically assigns weights (increasing the weight of the heading stability feature in strong crosswinds).

[0095] Specifically, the sustained climb rate anomaly is an aircraft performance anomaly judgment model based on multi-dimensional feature fusion. By integrating multi-source information such as aerodynamic performance, meteorological conditions, and route structure, it provides early warning of the performance degradation trend during the aircraft's climb process.

[0096] Specifically, the correction instructions include primary instructions, intermediate instructions, and advanced instructions; the primary instructions are to adjust the N1 limit value of the thrust management system, the intermediate instructions are to reconstruct the FMS vertical navigation profile, and the advanced instructions are to trigger a maintenance warning.

[0097] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device.

[0098] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0099] The program code included in the computer-readable medium can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, RF or the like, or any suitable combination thereof. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, and the programming language includes an object-oriented programming language such as Java, Smalltalk, C++, and also includes a conventional procedural programming language such as "C" language or similar programming language. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, utilizing an Internet service provider to connect through the Internet).

[0100] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An intelligent analysis method for aviation big data based on trajectory anomaly detection, characterized in that: include: S1: Acquire aircraft operating status data through the aircraft communication addressing and reporting system; obtain flight parameters from the quick access recorder, extract aircraft model aerodynamic parameters and corresponding horizontal navigation and vertical navigation deviation data from the navigation database in the flight management system; use the real-time processing engine to process the aircraft operating status data, flight parameters, aircraft model aerodynamic parameters, and deviation data in a batch-processing manner and unify them in time and space to obtain standardized data; S2: performing data cleaning on the standardized data, identifying and correcting abnormal points in the standardized data by combining the aircraft aerodynamic characteristics and the distribution of historical flight data, and interpolating missing values ​​for flight data signal loss in the standardized data; S3: Perform an aerodynamic-performance correlation analysis based on the cleaned data. Quantify the impact of parameter deviations on climb performance using a dynamic response model to obtain a strong correlation coefficient between key flight parameters and climb altitude anomalies. A climb rate attenuation prediction model is then used to couple this strong correlation coefficient with meteorological data and route structure data to obtain an aircraft climb performance anomaly prediction result. S4: Generate a correction instruction according to the sustained climb rate anomaly in the performance anomaly prediction result.

2. The method according to claim 1, characterized in that The method for integrated flow and batch processing is: S101: Establishing a hybrid computing architecture consisting of a distributed message queue and a micro-batch processing channel, accessing the streaming data stream generated by the aircraft communication addressing and reporting system in real time through the distributed message queue, and establishing the micro-batch processing channel to load batch flight data in parallel; S102: Using event time semantics to define data validity windows, applying a sliding time window function to the flow data of the aircraft communication addressing and reporting system to perform aggregation calculations, and resampling the flight data in segments based on the flight phase identifier; S103: By establishing an aircraft dynamic status register, operating parameters including airspeed, heading angle, and engine thrust are continuously updated. When the complex event processing engine detects a flight profile switching event, it automatically activates the batch processing layer to retrieve the pre-stored aircraft aerodynamic parameter library and perform parameter relevance verification. S104: Using the protocol buffer, perform structured deserialization processing on the aircraft communication addressing and reporting system message; Based on the civil aviation standard flight data frame structure, binary parsing is performed on the fast access recorder parameter set to construct a four-dimensional feature standardized data cube including timestamp, three-dimensional spatial coordinates, flight phase status and dynamic operation parameters.

3. The method according to claim 1, characterized in that The time-space reference unification method includes: The timestamps embedded in the messages are extracted from the streaming data of the aircraft communication addressing and reporting system; the absolute flight time recorded in the batch data of the fast access recorder is parsed; the two types of time sources are calibrated using the Network Time Protocol and uniformly converted into the Coordinated Universal Time standard time series; a dual-streamline mechanism based on event time is established to dynamically compensate for the transmission delay between streaming real-time data and batch historical data, and the real-time data stream and historical batch data are aligned along the time dimension through time window correlation technology. The longitude and latitude coordinates in the position report of the aircraft communication addressing and reporting system are converted into the Earth-centered Earth-fixed three-dimensional rectangular coordinate system. The inertial navigation system data recorded by the fast access recorder are normalized to the coordinate system. The reference deviation between different navigation sources is eliminated through the coordinate rotation matrix to generate a unified spatial trajectory sequence. Resample the fast access recorder parameters in the time dimension under a unified time and space reference, and implement piecewise linear interpolation based on the flight segment profile in the spatial dimension in combination with the flight phase identifier; The Kalman filter fuses multi-source position data to construct a multi-dimensional state vector of longitude, latitude, barometric altitude, and GNSS altitude. Trajectory continuity is predicted based on the aircraft's kinematic model. When a coordinate system jump or timestamp breakpoint is detected, dead reckoning based on heading angle and ground speed is triggered for data compensation. The unified spatiotemporal benchmark data is encoded into a composite key structure, and a three-level spatiotemporal index is established in the distributed storage system.

4. The method according to claim 1, wherein The method for identifying and correcting abnormal points in the standardized data based on the distribution of aircraft aerodynamic characteristics and flight segment history data is as follows: S201: Based on the lift coefficient, drag coefficient, and engine thrust curves in the aircraft aerodynamic parameter library, combined with the current flight altitude and Mach number, the aerodynamic performance envelope boundaries are dynamically generated. Envelope analysis is performed on the angle of attack, airspeed, and thrust-to-weight ratio parameters in the standardized data. If a point exceeds the aerodynamic performance envelope threshold for three consecutive sampling periods, it is marked as an aerodynamic characteristic abnormality point. S202: Filter historical flight data for the same route and aircraft configuration from the flight segment history database to construct a benchmark dataset; use a dynamic time warping algorithm to achieve nonlinear alignment of the flight phase time axis, eliminating differences in flight duration, phase transition rhythm, and other aspects between different flight segments; use a density-based ordered clustering algorithm to perform feature analysis on multidimensional flight parameters and construct multidimensional distribution features of the flight segment parameters; calculate the Mahalanobis distance of the current standardized parameter based on the covariance matrix of the cluster feature space; trigger a flight segment distribution anomaly warning when the Mahalanobis distance of the observed value deviates from the cluster center by more than three standard deviations; S203: For aerodynamic characteristic anomalies, an extended Kalman filter is used to perform state estimation, and the redundant measurement values ​​of the inertial navigation system and the atmospheric data system are integrated to generate a residual sequence, triggering the reverse deduction correction based on the aerodynamic equations. For flight segment distribution anomalies, the meteorological radar data and the monitoring data of the automatic dependent surveillance-broadcast system are correlated to build a multi-source verification channel. If the external data source verifies that the anomaly persists, the original data is retained and a credibility label is added.

5. The method according to claim 1, wherein The missing value interpolation method is: Based on the flight phase identifier and spatiotemporal reference data, the flight phase in which the rapid access recorder signal is lost is identified, and a differentiated interpolation strategy is selected according to the phase characteristics. For data windows where the duration of continuous missing data is shorter than the duration of stability, interpolation is performed based on adjacent valid data points, and the smoothing coefficient is dynamically calculated through the sliding window to adapt to the data fluctuation characteristics; For data windows whose continuous missing duration exceeds the stability duration and are in the stable cruising stage, a pattern matching method based on historical flight data is used to retrieve historical data segments that match the flight stages before and after the current missing data window from the flight segment history database, extract the time series of corresponding parameters in the historical data segments, and match the historical data segments that are most similar to the missing data window through the time series similarity measurement algorithm. Based on this, linear interpolation or spline interpolation method is used to interpolate missing values.

6. The method according to claim 1, characterized in that The aerodynamic-performance correlation analysis method is: S301: Construct an aerodynamic differential equation model based on the aircraft model aerodynamic parameter library, associate the lift coefficient, drag coefficient with flight state parameters, and establish a dynamic response relationship: Where T is the engine thrust, D is the aerodynamic drag, L is the lift, W is the weight of the aircraft, γ is the climb angle; α is the angle of attack, Ma is the Mach number, and H is the altitude; The partial derivatives of aerodynamic parameters on climb rate are calculated through parameter sensitivity analysis to quantify the influence of each parameter deviation on performance. S302: Using a state-space model to characterize the longitudinal motion of the aircraft, using QAR parameter deviations (e.g., N1 rotation speed deviation ΔN1, airspeed deviation ΔV) as input disturbances and climb altitude deviation ΔH as output responses; S303: Design a sliding time test to dynamically identify the causal relationship strength between parameter deviation and performance anomaly, and screen out the key QAR parameter set with a strong correlation coefficient (|r|>0.8).

7. The method according to claim 1, characterized in that The dynamic response model integrates the engine performance attenuation factor with the wing surface pressure distribution data, and calculates the aerodynamic efficiency loss through finite element analysis. It also introduces meteorological data as boundary conditions and uses computational fluid dynamics transient simulation to evaluate the disturbance effect of the external environment on the climb performance.

8. The method according to claim 1, characterized in that The calculation method of the climb rate attenuation prediction model is: A spatiotemporal graph convolutional network is established. It uses the four-dimensional features of a standardized data cube as input to construct a graph structure containing aerodynamic topological relationships. The nodes represent flight state parameters, and the edge weights are determined by the aerodynamic coupling coefficient. Train a multi-task learning model to simultaneously predict the climb rate attenuation trend and anomaly level. The loss function is designed as: Among them, λ1, λ2 are task weight coefficients, MSE is mean square error loss, CrossEntropy is cross entropy loss, y RoC is the target climb rate value of the sample, Represents the target climb rate value predicted by the model, y class is the actual abnormality level label, represents the probability distribution of anomaly levels predicted by the model; An online incremental learning engine is deployed to dynamically update the aerodynamic-performance correlation matrix when new flight segment data is detected, and random forest feature importance ranking is used to correct strong correlation coefficients in real time.

9. The method according to claim 1, characterized in that The sustained climb rate anomaly is an aircraft performance anomaly determination model based on multi-dimensional feature fusion. By integrating multi-source information such as aerodynamic performance, meteorological conditions, and route structure, it provides early warning of performance degradation trends during an aircraft's climb.

10. The method according to claim 1, characterized in that The correction instructions include primary instructions, intermediate instructions, and advanced instructions; the primary instructions are to adjust the N1 limit value of the thrust management system, the intermediate instructions are to reconstruct the FMS vertical navigation profile, and the advanced instructions are to trigger a maintenance warning.

Images