CN-122016114-A - Data-driven unmanned aerial vehicle wing dynamic load anomaly identification method and system

CN122016114ACN 122016114 ACN122016114 ACN 122016114ACN-122016114-A

Abstract

The invention discloses a data-driven unmanned aerial vehicle wing dynamic load anomaly identification method and system, and relates to the technical field of data processing. The method comprises the steps of obtaining reference load interval distribution and monitoring flight load distribution in real time. And when the monitored flight load distribution exceeds the reference load interval distribution, calling a historical environment data sequence. And predicting the abnormal state of the target cargo based on the historical environment data sequence to obtain the probability distribution of the abnormal state. And combining the historical environment data sequence and the abnormal state probability distribution to generate an influence load distribution, calculating the similarity between the distribution and the actual flight load distribution, and calculating the wing load abnormality rate to be used as a load abnormality identification result. According to the invention, through a data driving mode, the load abnormality caused by environmental interference, cargo state change and unmanned aerial vehicle body faults is effectively distinguished, the identification accuracy and diagnosis pertinence are improved, and a reliable technical support is provided for unmanned aerial vehicle freight safety.

Inventors

Ni Yingge

Assignees

西安航空学院

Dates

Publication Date: 20260512
Application Date: 20260123

Claims (10)

1. The data-driven unmanned aerial vehicle wing dynamic load anomaly identification method is characterized by comprising the following steps of: Acquiring reference load interval distribution of a rear wing of the unmanned plane carrying the target cargo, and monitoring the flight load distribution through a sensor in the flight process; when the flight load distribution exceeds the reference load interval distribution, a historical environment data sequence is called; According to the historical environment data sequence, carrying out abnormal state prediction of the target goods to obtain abnormal state probability distribution; and analyzing and obtaining influence load distribution according to the historical environment data sequence and the abnormal state probability distribution, calculating the similarity with the flight load distribution, and calculating the wing load abnormal rate to be used as a load abnormal identification result.
2. The method for identifying dynamic load anomalies of a data-driven unmanned aerial vehicle wing according to claim 1, wherein obtaining a reference load interval distribution of the unmanned aerial vehicle wing after carrying the target cargo, during flight, monitoring the flight load distribution by a sensor, comprises: according to load test data of unmanned aerial vehicle flight, processing and obtaining reference load interval distribution of the unmanned aerial vehicle carrying target cargo rear wing; During the flight of the unmanned aerial vehicle, the flight load distribution is monitored and acquired through the wing-configured sensors, wherein the flight load distribution comprises loads of a plurality of wing areas.
3. The method for identifying dynamic load anomalies of a wing of an unmanned aerial vehicle driven by data according to claim 2, wherein the step of processing and acquiring reference load interval distribution of the wing of the unmanned aerial vehicle on which the target cargo is carried according to load test data of unmanned aerial vehicle flight comprises the steps of: acquiring a test load distribution set according to wing loads tested in a normal flight process after the similar unmanned aerial vehicle carries the similar target cargoes; and constructing reference load interval distribution according to the end point values of the wing areas in the test load distribution set.
4. The method of claim 1, wherein retrieving a historical environmental data sequence when the flight load profile exceeds the baseline load interval profile comprises: judging whether the load of any wing area in the flying load distribution exceeds a corresponding reference load zone or not; If yes, processing to acquire a historical environment data sequence within a past preset time range, and if not, continuing to monitor the flight load distribution, wherein the historical environment data sequence comprises environment airflow data within the past preset time range.
5. The method for identifying dynamic load anomalies of a data-driven unmanned aerial vehicle wing according to claim 1, wherein the predicting the anomaly state of the target cargo according to the historical environmental data sequence, obtaining an anomaly state probability distribution, comprises: Acquiring a cargo state predictor, wherein the cargo state predictor comprises a plurality of cargo abnormality prediction branches corresponding to a plurality of cargo abnormality states; And inputting the historical environment data sequence into the cargo state predictor, outputting the occurrence probability of various cargo abnormal states, and obtaining the abnormal state probability distribution.
6. The method for identifying dynamic load anomalies of a data-driven unmanned aerial vehicle wing of claim 5, wherein obtaining the cargo state predictor comprises: Collecting a sample flight environment data sequence set according to test data of the same kind of cargo carried by the same kind of unmanned aerial vehicle, obtaining probabilities of various cargo abnormal states of the cargo under different sample flight environment data sequences, and labeling to obtain a plurality of sample abnormal state probability sets; constructing a plurality of cargo anomaly prediction branches corresponding to a plurality of cargo anomaly states based on machine learning; and taking the sample flight environment data sequence set as training input, respectively taking the plurality of sample abnormal state probability sets as supervision labels, respectively performing supervision training and testing on a plurality of cargo abnormal prediction branches, and obtaining a cargo state predictor after convergence.
7. The method for identifying dynamic load anomalies of a data-driven unmanned aerial vehicle wing according to claim 1, wherein analyzing and obtaining an impact load distribution set according to the historical environmental data sequence and the anomaly state probability distribution, calculating similarity with the flight load distribution, and calculating a wing load anomaly rate as a load anomaly identification result comprises: Acquiring an influence load predictor, wherein the influence load predictor comprises a plurality of influence load prediction branches corresponding to various cargo abnormal states, and each influence load prediction branch is obtained by training a sample environment data sequence set and a sample influence load distribution set which are tested under different cargo abnormal states; inputting the historical environment data sequence into the influence load predictor, and outputting to obtain a plurality of predicted influence load distributions; according to the abnormal state probability distribution, calculating the plurality of predicted influence load distribution to obtain influence load distribution; and calculating the similarity of the influence load distribution and the flying load distribution, and calculating the abnormal rate of the wing load as a load abnormal identification result.
8. The method for identifying dynamic load anomalies of a data-driven unmanned aerial vehicle wing according to claim 7, wherein computing the plurality of predicted impact load distributions according to the anomaly state probability distribution to obtain an impact load distribution comprises: Distributing a plurality of state weights according to the occurrence probability of the abnormal states of a plurality of cargoes in the abnormal state probability distribution; and carrying out weighted calculation on the predicted influence loads of the same wing area in the predicted influence load distribution by adopting a plurality of state weights to obtain influence load distribution.
9. The method for identifying dynamic load anomalies of a data-driven unmanned aerial vehicle wing according to claim 7, wherein calculating the similarity of the impact load distribution and the flight load distribution and calculating a wing load anomaly rate as a result of load anomaly identification comprises: calculating a load similarity of the impact load distribution and the flight load distribution; And according to the load similarity, calculating to obtain the wing load abnormality rate as a load abnormality identification result.
10. A data-driven unmanned aerial vehicle wing dynamic load anomaly identification system for performing the data-driven unmanned aerial vehicle wing dynamic load anomaly identification method of any one of claims 1-9, comprising: The load monitoring module is used for acquiring reference load interval distribution of the rear wing of the unmanned aerial vehicle carrying the target cargo, and monitoring the flight load distribution through the sensor in the flight process; the data calling module is used for calling a historical environment data sequence when the flight load distribution exceeds the reference load interval distribution; The anomaly prediction module is used for predicting the anomaly state of the target cargo according to the historical environment data sequence to obtain anomaly state probability distribution; The anomaly identification module is used for analyzing and obtaining influence load distribution according to the historical environment data sequence and the anomaly state probability distribution, calculating similarity with the flight load distribution, and calculating the wing load anomaly rate to be used as a load anomaly identification result.

Description

Data-driven unmanned aerial vehicle wing dynamic load anomaly identification method and system Technical Field The invention relates to the technical field of data processing, in particular to a data-driven unmanned aerial vehicle wing dynamic load anomaly identification method and system. Background When the unmanned aerial vehicle carries cargoes to execute tasks, the wing is used as a key load-carrying structure, the dynamic load state of the unmanned aerial vehicle directly relates to the stability and safety of the flight, and the wing can be subjected to load abnormality due to gusts, structural damage or cargo state change and the like, and if the unmanned aerial vehicle is not recognized in time, the flight attitude is easy to be out of control, and even serious accidents are caused. Currently, the recognition technology for the load abnormality of the unmanned aerial vehicle is mostly dependent on a preset fixed threshold value or a judgment method based on a simple physical model. Load data of the wing are collected through the sensor and compared with a safety threshold value, and if the load data exceeds the safety threshold value, the load data is judged to be abnormal. However, in actual flight, the load signal itself may be disturbed by various complex factors, such as abrupt gusts, unstable air flows, and noise of the sensor itself. Various interference signals are easily misjudged by the existing method to be abnormal in load caused by structural damage or control faults, so that false alarms are frequently generated by the system, and the accuracy and the reliability of identification are reduced. Meanwhile, the existing method generally only can make binary judgment of abnormality and normal, lacks analysis and quantitative evaluation capability for abnormality reasons, cannot accurately distinguish whether the abnormality is caused by external environment interference or structural problems of the body, and is difficult to provide effective subsequent decision support. Disclosure of Invention Aiming at the technical problems that in the prior art, load abnormality identification is easy to be interfered by environment, low in accuracy and incapable of effectively distinguishing abnormality sources, the invention provides a data-driven unmanned aerial vehicle wing dynamic load abnormality identification method and system. The technical scheme for solving the technical problems is as follows: In a first aspect, the present invention provides a method for identifying dynamic load anomalies of a data-driven unmanned aerial vehicle wing, comprising: Acquiring reference load interval distribution of a rear wing of the unmanned plane carrying the target cargo, and monitoring the flight load distribution through a sensor in the flight process; when the flight load distribution exceeds the reference load interval distribution, a historical environment data sequence is called; According to the historical environment data sequence, carrying out abnormal state prediction of the target goods to obtain abnormal state probability distribution; and analyzing and obtaining influence load distribution according to the historical environment data sequence and the abnormal state probability distribution, calculating the similarity with the flight load distribution, and calculating the wing load abnormal rate to be used as a load abnormal identification result. In a second aspect, the present invention provides a data-driven unmanned aerial vehicle wing dynamic load anomaly identification system, comprising: The load monitoring module is used for acquiring reference load interval distribution of the rear wing of the unmanned aerial vehicle carrying the target cargo, and monitoring the flight load distribution through the sensor in the flight process; the data calling module is used for calling a historical environment data sequence when the flight load distribution exceeds the reference load interval distribution; The anomaly prediction module is used for predicting the anomaly state of the target cargo according to the historical environment data sequence to obtain anomaly state probability distribution; The anomaly identification module is used for analyzing and obtaining influence load distribution according to the historical environment data sequence and the anomaly state probability distribution, calculating similarity with the flight load distribution, and calculating the wing load anomaly rate to be used as a load anomaly identification result. The beneficial effects of the invention are as follows: Compared with the prior art, the method has the advantages that the dynamic reference load interval distribution is established to replace the traditional fixed threshold value method, the adaptability to different flight conditions is enhanced, and misjudgment caused by the change of the conditions is reduced. And secondly, a cargo abnormal state prediction link based on historical environment data is int