CN-121997058-A - Unmanned aerial vehicle state parameter prediction method and system based on operation condition data cluster division

CN121997058ACN 121997058 ACN121997058 ACN 121997058ACN-121997058-A

Abstract

An unmanned plane state parameter prediction method and system based on operation condition data cluster division belongs to the technical field of aircrafts, and solves the problem that the accuracy of a prediction model is reduced due to complex unmanned plane data characteristics and new operation conditions under multiple operation conditions. The method comprises the steps of collecting flight data of the unmanned aerial vehicle, preprocessing the flight data, dividing the preprocessed flight data to obtain a flight data set, establishing a reference prediction model library, obtaining an optimal similarity result matrix according to data to be detected of the unmanned aerial vehicle by adopting an optimal prediction model matching method based on similarity measurement, selecting a reference prediction model according to optimal similarity matching, adjusting training parameters of the reference prediction model by adopting a domain adaptive algorithm, updating the reference prediction model, and predicting state parameters of the unmanned aerial vehicle by adopting the updated reference prediction model. The method is suitable for the unmanned aerial vehicle state monitoring scene.

Inventors

WANG BENKUAN
SU ZHOU
WANG YUAN
WANG NA
LIU DATONG

Assignees

哈尔滨工业大学

Dates

Publication Date: 20260508
Application Date: 20241108

Claims (10)

1. An unmanned aerial vehicle state parameter prediction method based on operation condition data cluster division is characterized by comprising the following steps: s1, acquiring flight data of an unmanned aerial vehicle, preprocessing, and dividing the preprocessed flight data to obtain a flight data set; S2, establishing a reference prediction model library according to the flight data set; s3, according to the data to be tested of the unmanned aerial vehicle, an optimal prediction model matching method based on similarity measurement is adopted to obtain an optimal similarity result matrix, and according to optimal similarity matching, a reference prediction model is selected; s4, adjusting training parameters of the reference prediction model by adopting a domain self-adaptive algorithm, and updating the reference prediction model; s5, predicting the state parameters of the unmanned aerial vehicle by adopting the updated reference prediction model.
2. The unmanned aerial vehicle state parameter prediction method based on the operation condition data cluster division of claim 1, wherein the preprocessing comprises data cleaning, data alignment, data standardization and data reconstruction of the flight data.
3. The unmanned aerial vehicle state parameter prediction method based on the operation condition data cluster division of claim 1, wherein the division of the preprocessed flight data is to divide the processed data into data clusters by adopting a DTW algorithm.
4. The unmanned aerial vehicle state parameter prediction method based on operating condition data cluster division according to claim 3, wherein the data cluster division comprises: calculating to obtain the similarity of the unmanned aerial vehicle running state observation variables DTW of different frames; traversing different frames and calculating the DTW similarity thereof to form a similarity matrix; calculating minimum similarity values of different frames according to the similarity matrix rows and updating a coefficient matrix; And iterating and merging the frames until all the data clusters are partitioned.
5. The unmanned aerial vehicle state parameter prediction method based on the operation condition data cluster division of claim 1, wherein the reference prediction model library is constructed by 1 DCNN.
6. The unmanned aerial vehicle state parameter prediction method based on the operation condition data cluster division according to claim 1, wherein the optimal prediction model matching method comprises the following steps: calculating the DTW similarity between the target domain data set and the flight data set; sequentially calculating the DTW similarity between each flight data set and the target domain data set to obtain a similarity result matrix; and selecting the flight data set with the highest similarity degree for matching the target domain data set.
7. The unmanned aerial vehicle state parameter prediction method based on the operation condition data cluster division according to claim 1, wherein the S4 comprises: Setting a domain self-adaptive algorithm to adjust model training parameters; updating a prediction model of the key parameters of the running state, reserving parameters of a convolution layer and a pooling layer of a source domain training model, and updating a flattening layer and two full-connection layers.
8. An unmanned aerial vehicle state parameter prediction system based on operating condition data cluster division, the system comprising: The data processing module is used for acquiring flight data of the unmanned aerial vehicle, preprocessing the flight data, and dividing the preprocessed flight data to obtain a flight data set; The model building module is used for building a reference prediction model library according to the flight data set; The optimal matching module is used for obtaining an optimal similarity result matrix by adopting an optimal prediction model matching method based on similarity measurement according to the data to be detected of the unmanned aerial vehicle, and selecting a reference prediction model according to the optimal similarity matching; The model updating module is used for adjusting training parameters of the reference prediction model by adopting a domain self-adaptive algorithm and updating the reference prediction model; And the parameter prediction module is used for predicting the state parameters of the unmanned aerial vehicle by adopting the updated reference prediction model.
9. Computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes an unmanned aerial vehicle state parameter prediction method based on operating condition data cluster division according to any one of claims 1-7.
10. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program which, when executed by a processor, implements the steps of an unmanned aerial vehicle state parameter prediction method based on operating condition data cluster division according to any one of claims 1-7.

Description

Unmanned aerial vehicle state parameter prediction method and system based on operation condition data cluster division Technical Field The invention belongs to the technical field of aircrafts, and particularly relates to an aircraft state monitoring technology. Background Various systems of the unmanned aerial vehicle can collect and record a large amount of flight data in the flight process, and the flight data contains unmanned aerial vehicle running state information. The unmanned aerial vehicle state monitoring technology can utilize the state information to judge whether the unmanned aerial vehicle deviates from a given running state, and effectively discover potential safety hazards of the unmanned aerial vehicle in time, supports adjustment of a subsequent decision task plan, and is an important technical measure for protecting flight safety of the unmanned aerial vehicle. The prediction model can be trained according to unmanned aerial vehicle flight data in normal expectation based on the prediction method, and the actual running state of the unmanned aerial vehicle is monitored through accurate prediction of key parameters. However, the actual operation condition of the unmanned aerial vehicle is not constant, the diversity and the dynamics of the operation condition necessarily lead to more complex characteristics of the collected flight data, and the correlation between the change of the distribution characteristic of the flight data and the multidimensional data is changed, so that the operation state key parameter prediction model is difficult to construct accurately. If a single model is built for the whole operation process according to a global modeling method, a plurality of working condition characteristics are averaged, important local data characteristic information is extremely easy to lose, and if accurate models are built for different working conditions according to a multi-model method, the method is an important challenge for data division and model classification, and a large amount of storage space and training time are needed for multi-model construction. Disclosure of Invention The invention provides an unmanned aerial vehicle state parameter prediction method and system based on operation condition data cluster division, and aims to solve the problem that the accuracy of a prediction model is reduced due to complex unmanned aerial vehicle data characteristics and the occurrence of a new condition under multiple operation conditions. The unmanned aerial vehicle state parameter prediction method based on the operation condition data cluster division provided by the invention comprises the following steps: s1, acquiring flight data of an unmanned aerial vehicle, preprocessing, and dividing the preprocessed flight data to obtain a flight data set; S2, establishing a reference prediction model library according to the flight data set; s3, according to the data to be tested of the unmanned aerial vehicle, an optimal prediction model matching method based on similarity measurement is adopted to obtain an optimal similarity result matrix, and according to optimal similarity matching, a reference prediction model is selected; s4, adjusting training parameters of the reference prediction model by adopting a domain self-adaptive algorithm, and updating the reference prediction model; s5, predicting the state parameters of the unmanned aerial vehicle by adopting the updated reference prediction model. Still further, a preferred embodiment is provided wherein the preprocessing includes data cleansing, data alignment, data normalization and data reconstruction of the flight data. Furthermore, a preferable scheme is provided, wherein the division of the preprocessed flight data is to divide the processed data into data clusters by adopting a DTW algorithm. Still further, a preferred solution is provided wherein the data cluster partitioning comprises: calculating to obtain the similarity of the unmanned aerial vehicle running state observation variables DTW of different frames; traversing different frames and calculating the DTW similarity thereof to form a similarity matrix; calculating minimum similarity values of different frames according to the similarity matrix rows and updating a coefficient matrix; And iterating and merging the frames until all the data clusters are partitioned. Still further, a preferred solution is provided wherein the library of reference predictive models is constructed using 1 DCNN. Still further, the optimal prediction model matching method comprises the following steps: calculating the DTW similarity between the target domain data set and the flight data set; sequentially calculating the DTW similarity between each flight data set and the target domain data set to obtain a similarity result matrix; and selecting the flight data set with the highest similarity degree for matching the target domain data set. Still further, a preferred embodiment is provided,