Abnormal Detection Algorithm Research And System Design Of Aerodynamic Data

Aerodynamic data is the key of Aircraft Aerodynamics simulation,which involves the whole process of aircraft design and manufacturing.In the process of calculation and collection,abnormal data may be generated.If the abnormal data is used for aerodynamics simulation directly,it may cause model distortion,wrong results and inestimable losses.Anomaly detection of aerodynamic data can detect anomalies in advance,which is conducive to the design and manufacture of aircraft.The traditional method is to rely on professional manual identification.However,due to the large volume,high dimension and complex relationship of aerodynamic data,manual identification of anomalies is time-consuming and error prone,which seriously restricts the progress of related research.At present,there are few researches on abnormal detection of aerodynamic data.This paper analyzes the characteristics of aerodynamic data,investigates the abnormal detection algorithm,studies the abnormal detection algorithm suitable for aerodynamic data,and designs and implements the detection system based on the algorithm,which can realize fast and efficient detection of abnormal in aerodynamic data and reduce the workload of researchers,It has a positive significance for the research of aircraft.The research work and achievements of this paper are as follows(1)By analyzing the characteristics of aerodynamic data,it is found that the high dimension of aerodynamic data,the coexistence of functional and non-functional relationships,and the imbalance of data become the difficulties of anomaly detection.It is necessary to combine the characteristics of aerodynamic data with the detection requirements to study the suitable detection methods.In this paper,the widely used anomaly detection methods are investigated,the principle of the algorithm is described in detail,and the feasibility of applying it to the anomaly detection of aerodynamic data is analyzed.(2)In this paper,the robustness and efficiency of four kinds of robust regression techniques are compared and analyzed,and the ISVD-Fast LTS algorithm based on robust regression is proposed.The algorithm takes the clear mathematical and physical relationship between the detected objects as the breakthrough point,uses Fast LTS with high collapse value to resist the interference of outliers,and uses ISVD to solve the large-scale matrix solving problem of Fast LTS,and quickly establishes the regression model in line with most of the data and carries out anomaly detection.(3)The SMESM-EIF algorithm is proposed.The variables related to the anomaly are selected by SMESM,and the isolation tree is established by EIF.By generating the random hyperplane isolation data,the abnormal data will be isolated earlier.The algorithm has linear time complexity,and can complete anomaly detection for high-dimensional complex data.(4)In order to make the algorithm more simple and easy to use,this paper designs and implements a pneumatic data anomaly detection system based on the research algorithm,and introduces the system development in detail from four aspects: system demand analysis,technology selection,algorithm implementation key technology and system design.The system provides a visualization module based on parallel coordinates and 3D scatter diagram for users to intuitively evaluate the test results and observe the abnormal details.Function and performance test results show that the system can meet the requirements.