Flight state analysis method based on test flight electronic and electrical data
By constructing a hierarchical detection architecture and a cross-bus data spatiotemporal alignment engine, the systemic and data silo problems in modern aircraft flight testing have been solved, enabling full-dimensional analysis and efficient fault diagnosis of electronic and electrical systems, and improving the accuracy and adaptability of flight status analysis.
Patent Information
- Application Number
- CN202512045564.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-31
- Publication Date
- 2026-05-01
AI Technical Summary
Existing analytical methods lack systematicity, have insufficient fault diagnosis capabilities, poor adaptability to condition monitoring, and suffer from data silos in modern aircraft flight testing, making it difficult to effectively handle highly integrated and networked airborne systems.
A hierarchical detection architecture is constructed, employing wavelet denoising, support vector machine classifier, and Bayesian network model, combined with dynamic threshold algorithm and cross-bus data spatiotemporal alignment engine, to achieve in-depth analysis at the signal, device, and system levels and cross-system anomaly identification.
It enables comprehensive analysis of flight test electronic and electrical data, improves the accuracy and efficiency of fault diagnosis, reduces false alarm and missed alarm rates, and enhances the accuracy and adaptability of aircraft-wide health status assessment.
Smart Images
Figure CN121959342A_ABST
Abstract
Description
A method for flight status analysis based on flight test electronic and electrical data Technical Field
[0001] This invention relates to the field of flight data processing technology, specifically to a flight status analysis method based on flight test electronic and electrical data. Background Technology
[0002] The massive amounts of electronic and electrical data generated during modern aircraft flight tests are crucial for system health monitoring and fault diagnosis. However, existing analytical methods have significant limitations when dealing with highly integrated and networked modern airborne systems.
[0003] First, the analysis architecture lacks a systematic hierarchy. Existing methods often handle signal layer anomalies or device layer states in isolation, making it difficult to effectively correlate and comprehensively diagnose the transient signal characteristics at the bottom layer, the device functional failures at the middle layer, and the system interaction mismatches at the top layer, resulting in insufficient insight into the root causes of complex coupled problems.
[0004] Secondly, the fault diagnosis capability is insufficient. Traditional methods based on rules or simple causal models are difficult to deal with scenarios of multiple concurrent faults and complex propagation paths caused by tight coupling in electronic and electrical systems, and cannot accurately perform probabilistic reasoning and root cause localization.
[0005] Furthermore, the condition monitoring system has poor adaptability. The commonly used static threshold alarm mechanism cannot adapt to the dynamic changes in operating conditions at various stages of flight (such as climb and maneuver), which easily generates a large number of false alarms or missed alarms, seriously affecting the monitoring effectiveness.
[0006] Finally, there is the problem of "data silos." Aircraft systems typically use multiple heterogeneous buses, and current technologies lack the ability for high-precision spatiotemporal synchronization and deep fusion, making it impossible to effectively detect cross-system correlation anomalies and limiting the accuracy of aircraft-wide health status assessments. Summary of the Invention
[0007] The purpose of this invention is to provide a flight status analysis method based on flight test electronic and electrical data. Through this method, a hierarchical detection architecture enables in-depth analysis across all dimensions, from signals and equipment to the system. A Bayesian network-based fault propagation model accurately locates the root causes of multiple concurrent faults. A dynamic threshold algorithm adapts to flight conditions, significantly reducing false alarms and missed alarms. A cross-bus data spatiotemporal alignment engine overcomes heterogeneous data barriers, achieving accurate identification of cross-system anomalies.
[0008] To achieve the above effects, the present invention provides the following technical solution: a flight status analysis method based on flight test electronic and electrical data, comprising the following steps:
[0009] S1. Construct a layered detection architecture, which includes a signal layer, a device layer, and a system layer, to improve the ability to analyze flight test data of electronic and electrical systems.
[0010] S2. Design a fault propagation model based on Bayesian networks to analyze the coupling relationship of subsystems and achieve accurate location of multiple concurrent faults.
[0011] S3. A dynamic threshold sliding window learning algorithm is adopted to adaptively adjust the alarm threshold to adapt to the dynamic working conditions of flight test.
[0012] S4. Develop a cross-bus data spatiotemporal alignment engine to achieve synchronization and fusion of heterogeneous bus data streams and identify cross-system abnormal events.
[0013] Furthermore, the construction of the hierarchical detection architecture includes:
[0014] At the signal layer, wavelet denoising technology is used to filter out high-frequency noise, and an adaptive feature extraction algorithm is combined to dynamically capture the transient features of voltage and current.
[0015] At the device layer, the operating status of the device is classified by a support vector machine classifier, and failure determination is performed for individual sensors and actuators.
[0016] At the system level, a multi-device state covariance analysis model is designed to calculate the deviation of signal interaction between subsystems in real time and identify functional coupling anomalies.
[0017] Furthermore, the fault propagation model is based on a Bayesian network to construct an electronic and electrical system topology model, defines component nodes and their signal transmission relationships, and calculates the propagation path of multiple concurrent faults through probabilistic reasoning.
[0018] Furthermore, the dynamic threshold sliding window learning algorithm uses the mean and variance of statistical feature parameters within a preset time window to dynamically adjust the alarm threshold based on the 3σ principle.
[0019] Furthermore, the cross-bus data spatiotemporal alignment engine synchronizes heterogeneous data streams of the avionics bus and power bus through timestamps, and uses an association rule mining algorithm to identify cross-system abnormal events.
[0020] Furthermore, the transient features dynamically captured by the signal layer adaptive feature extraction algorithm include voltage spikes and voltage dips.
[0021] Furthermore, the devices classified by the support vector machine classifier in the device layer include flight control computers and power modules.
[0022] Furthermore, the functional coupling anomalies identified at the system layer include deviations between avionics bus commands and power response delays.
[0023] Furthermore, the preset time window is 10 seconds.
[0024] Furthermore, the heterogeneous data streams include avionics bus data based on the ARINC 429 protocol and power bus data based on the CAN protocol.
[0025] This invention provides a flight status analysis method based on flight test electronic and electrical data, which has the following beneficial effects:
[0026] (1) This invention constructs a hierarchical detection architecture comprising a signal layer, a device layer, and a system layer, enabling systematic analysis of flight test electronic and electrical data from microscopic to macroscopic levels and from local to overall levels. At the signal layer, wavelet denoising and adaptive feature extraction are employed to effectively remove noise interference and accurately capture key transient features such as voltage spikes and drops, providing a high-quality data foundation for upper-level analysis. At the device layer, a support vector machine classifier is used to determine the status of key units such as flight control computers and power modules, achieving automated and intelligent identification of device-level failures. At the system layer, a covariance analysis model is used to quantify the interaction deviation between subsystems, effectively revealing functional coupling anomalies such as command and response delay mismatch, thereby comprehensively improving the observability and in-depth analysis capability of complex electronic and electrical system states.
[0027] (2) This invention effectively addresses the challenge of locating multiple concurrent faults in electronic and electrical systems caused by tight coupling by designing a fault propagation model based on Bayesian networks. This model constructs the system topology and defines the probabilistic dependencies between nodes, linking component failures, signal anomalies, and system-level functional failures. When an anomaly occurs, probabilistic reasoning can be used to reverse-calculate the most likely root cause and propagation path of the fault, thereby penetrating complex interactive phenomena and directly locating one or more concurrent fault points that cause systemic anomalies. This significantly improves the accuracy and efficiency of fault diagnosis, providing strong decision support for flight test safety analysis and troubleshooting.
[0028] (3) The dynamic threshold sliding window learning algorithm used in this invention successfully solves the problems of poor adaptability and high false alarm / missed alarm rates of traditional fixed threshold alarm methods under dynamic flight test conditions. This algorithm uses a sliding time window (e.g., 10 seconds) to statistically analyze the mean and variance of characteristic parameters in real time, and dynamically adjusts the alarm threshold based on the 3σ principle, enabling the threshold to adapt to the inherent statistical characteristics of data at different flight stages (e.g., climb, cruise, maneuver). This adaptive mechanism significantly improves the matching degree and reliability of condition monitoring and anomaly warning to the actual flight environment.
[0029] (4) The cross-bus data spatiotemporal alignment engine developed in this invention breaks through the technical barrier that makes it difficult to comprehensively utilize heterogeneous bus data streams due to differences in protocols and timing. By using timestamp synchronization technology, ARINC 429 avionics bus data and CAN power bus data are accurately aligned and fused, constructing a data view with a unified spatiotemporal reference across systems. On this basis, by using association rule mining algorithms to analyze the fused data, it is possible to discover and identify associated abnormal events that cannot be detected within a single bus data and that span multiple subsystems, thereby achieving a more comprehensive and accurate health status assessment and abnormal diagnosis from the system level to the entire aircraft level. Attached Figure Description
[0030] Figure 1 is a flowchart of a flight status analysis method system based on flight test electronic and electrical data according to the present invention. Detailed Implementation
[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.
[0032] This invention provides a technical solution, please refer to Figure 1:
[0033] Example 1:
[0034] A flight status analysis method based on flight test electronic and electrical data, characterized by the following steps:
[0035] Step 1: Construct a layered detection architecture
[0036] Construct a layered detection architecture that includes a signal layer, a device layer, and a system layer to systematically improve the ability to analyze flight test data of electronic and electrical systems.
[0037] Signal layer processing: First, wavelet denoising technology is used to filter the original voltage and current signals at this layer to remove high-frequency noise. Then, an adaptive feature extraction algorithm is combined to dynamically capture transient features in the signal, such as voltage spikes and voltage drops.
[0038] Device layer processing: At this layer, a pre-trained support vector machine classifier is used to classify and determine the operating status of key individual devices. For example, failure or abnormal status identification is performed on devices such as flight control computers, power modules, sensors, and actuators.
[0039] System-level processing: Design a multi-device state covariance analysis model to calculate the deviation of signal interactions between different subsystems in real time, thereby identifying anomalies caused by functional coupling. For example, identify coupling deviations where the power response delay exceeds the normal range after an avionics bus command is issued.
[0040] Step 2: Design a fault propagation model based on Bayesian networks
[0041] Design a fault propagation model based on Bayesian networks to analyze complex subsystem coupling relationships.
[0042] First, based on the actual physical connections and signal dependencies of the electronic and electrical system, its topology model is constructed, each component in the system is defined as a node, and the signal transmission relationship between them is clarified.
[0043] When an anomaly or fault is detected, the Bayesian network is used for probabilistic reasoning to calculate the probability of different fault nodes as root causes and their propagation paths, thereby achieving accurate localization of multiple concurrent faults.
[0044] Step 3: Employ a dynamic threshold sliding window learning algorithm
[0045] A dynamic threshold sliding window learning algorithm is adopted to adapt to the dynamically changing operating conditions during flight testing.
[0046] Set a preset time window (10 seconds in this embodiment) for specific feature parameters (such as the transient feature amplitude extracted in step one).
[0047] During the algorithm's operation, the mean and variance of the feature parameters are continuously calculated using this sliding window.
[0048] Based on the statistically obtained mean and variance, the alarm threshold of this feature parameter is dynamically calculated and adjusted according to the 3σ principle, thereby achieving threshold self-adaptation.
[0049] Step 4: Develop a cross-bus data spatiotemporal alignment engine
[0050] Develop a cross-bus data spatiotemporal alignment engine to achieve synchronization and fusion analysis of heterogeneous bus data.
[0051] First, timestamp synchronization is performed on heterogeneous data streams from different buses. In this embodiment, the heterogeneous data streams include avionics bus data based on the ARINC 429 protocol and power bus data based on the CAN protocol.
[0052] Based on time synchronization, the multi-source data streams are aligned and merged in the time domain to form a unified, time-consistent data view.
[0053] Finally, an association rule mining algorithm is used on the fused data to analyze the correlation between cross-system events, thereby identifying cross-system abnormal events that are difficult to detect with a single bus.
[0054] By sequentially executing steps one through four above, this embodiment achieves multi-level, automated flight status analysis of flight test electronic and electrical data, from signal processing, equipment diagnosis, system coupling analysis to cross-system anomaly identification.
[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A flight status analysis method based on flight test electronic and electrical data, characterized in that, Includes the following steps: S1. Construct a layered detection architecture, which includes a signal layer, a device layer, and a system layer, to improve the analytical capabilities of flight test data of electronic and electrical systems; S2. Design a fault propagation model based on Bayesian networks to analyze the coupling relationship of subsystems and achieve accurate location of multiple concurrent faults; S3. Adopt a dynamic threshold sliding window learning algorithm to adaptively adjust alarm thresholds to adapt to dynamic flight test conditions. S4. Develop a cross-bus data spatiotemporal alignment engine to achieve synchronization and fusion of heterogeneous bus data streams and identify cross-system abnormal events.
2. The flight status analysis method based on flight test electronic and electrical data according to claim 1, characterized in that, The constructed hierarchical detection architecture includes: at the signal layer, wavelet denoising technology is used to filter out high-frequency noise, and adaptive feature extraction algorithm is combined to dynamically capture transient features of voltage and current; at the device layer, a support vector machine classifier is used to classify the device operating status and to determine the failure of individual sensors and actuators; at the system layer, a multi-device state covariance analysis model is designed to calculate the deviation of signal interaction between subsystems in real time and identify functional coupling anomalies.
3. The flight status analysis method based on flight test electronic and electrical data according to claim 1, characterized in that, The fault propagation model is based on a Bayesian network to construct an electronic and electrical system topology model, defines component nodes and their signal transmission relationships, and calculates the propagation path of multiple concurrent faults through probabilistic reasoning.
4. The flight status analysis method based on flight test electronic and electrical data according to claim 1, characterized in that, The dynamic threshold sliding window learning algorithm uses a preset time window to statistically analyze the mean and variance of feature parameters and dynamically adjusts the alarm threshold based on the 3σ principle.
5. The flight status analysis method based on flight test electronic and electrical data according to claim 1, characterized in that, The cross-bus data spatiotemporal alignment engine synchronizes heterogeneous data streams of the avionics bus and power bus through timestamps and uses an association rule mining algorithm to identify cross-system abnormal events.
6. The flight status analysis method based on flight test electronic and electrical data according to claim 2, characterized in that, The transient features dynamically captured by the adaptive feature extraction algorithm of the signal layer include voltage spikes and voltage drops.
7. The flight status analysis method based on flight test electronic and electrical data according to claim 2, characterized in that, The devices classified by the support vector machine classifier in the device layer include flight control computers and power modules.
8. The flight status analysis method based on flight test electronic and electrical data according to claim 2, characterized in that, The functional coupling anomalies identified at the system layer include deviations between avionics bus commands and power response delays.
9. The flight status analysis method based on flight test electronic and electrical data according to claim 4, characterized in that, The preset time window is 10 seconds.
10. A flight status analysis method based on flight test electronic and electrical data according to claim 5, characterized in that, The heterogeneous data streams include avionics bus data based on the ARINC 429 protocol and power bus data based on the CAN protocol.