The Research Of Fault Detection Method For UAV's Navigation System

With the rapid development of UAV,the fault detection technology for the UAV's navigation system becomes very important.Toimprove the safety and reliability of the UAV,it can discovery faults by means of fault diagnosis.Compare to other aim,UAV's navigation system has fewer data of fault and it is a typical small sample learning problems.So it is worth studying about how to use limited fault sample size to get sufficient fault diagnosis.This thesis application the Support Vector Machine(SVM)technology to the fault detection of the UAV's navigation system,so it can complete the fault detection task by using limited training data.In this thesis,research on the fault detection method for the UAV's navigation system,the main work is as follows:First:Based on the compositions of the navigation system of a certain kind of UAV,this thesis analyses the navigation sensor and formulates the corresponding error model and fault model.This paves the basis for the construction of classifier.Second:Traditional fault diagnosis with limited number of samples tends to converge to a local optimal solution.In view of this,this thesis introduces the support vector machine method in UAV navigation system fault detection.Based on the practical situation of the navigation system and the multiple classifiers,this thesis applies the binary tree decomposition method to classify the fault state of the navigation system and trains the classifier with UAV navigation sensor operation data,which shortens classification time and improves accuracy.Third:Parameter selection plays an important role in SVM algorithm.Improper parameters can cause considerable error in pattern classification.This thesis makes some improvements to the traditional SVM algorithm and proposes the ant colony optimization method for the determination of key parameters,which is more reasonable and pertinent to the problem.Forth:In accordance with the requirements of experimental verification,this thesis designs the corresponding hardware and software,sets up the semi-physical experimental verification platform and tests the software compatibility.This thesis also designs fault simulator and fault scenario.The result of verification experiments proves the rationality and effectiveness of the fault detection method.