Research Of The Diagnosis Methods For Flighter Control Surface

The main work of this thesis is fault diagnosis for flighter control surface.Three diagnosis methods,rough neural networks,wavelet analysis and fractal geometry methods are emphasized,then these methods are validated by simulations.First,the thesis summarizes the states of research about fault diagnosis for flighter control surface,points out the disadvantages.Then,the way of this thesis is ascertained,that is,the thesis should be emphasized in the model-independent methods.A Multilayer fault-tolerance neural networks fault diagnosis baed on rough sets method is presented after we summarize rough neural networks in existence.A fault diagnosis method using wavelet analysis to detect signal singularity and extract fault symptom,and then using neural networks to classify is proposed.The thesis describes a novel fractal fault diagnosis method based on rough sets. Mathematic models about fault classification and discrimination are established by means of computing box-counting dimensions of fault signals.Finally,a emulational diagnosis system is devised by means of VC++ transferring MATLAB engine.The results of simulations indicate the validity of aforementioned three methods.