Navigation Sensors Fault Diagnosis Based On Wavelet Multi-scaled Entropy

Navigation devices have become the necessary part of the modern aerospace, marinetechnology, as to the multi-sensor integrated navigation system while providing navigationdata, is also fully utilized in the field of military and industrial. When a component failure,unpredictable consequences will occur, the affection ranging from affecting production,machine crash and even endanger the national defense. Therefore, the navigation sensor faultdiagnosis is very necessary. The paper used the navigation sensors components gyroscope,GPS systems as the analysis object, put forward a new artificial intelligence fault diagnosistechnology which conclude wavelet transform, information entropy technology, complexdegree measure and neural network technology. By analyzing the working principle andfailure causes of the gyroscope and GPS system, summed up the fault model, as to thesefailure characteristics of the navigation sensor component, put forward algorithm based onwavelet multi-scaled entropy.In order to overcome the shortcomings of the fault diagnosis requires establish the signalmathematical model, the paper used wavelet transform method, detailed introduced theanalysis ability in time and frequency domain, and use its multi-resolution analysis feature toanalysis the signal in different frequency. Through the simulate analysis, analysis thegyroscope in different scaled, concluded that wavelet transform can analysis the part andwhole feature of the signal through change the scale. Used the analysis and reconstructedtheory of wavelet transform and the wavelet threshold noise reduction principle topreparatory process the signal. In order to make up the shortcoming in reflecting theinformation of the signal, introduced the concept of information entropy, detailed introducemany kinds of entropy in different uses, as to the fault feature of the gyroscope, combine themulti-resolution theory of wavelet transform and the concept of information entropy, definedthe concept of wavelet multi-scaled entropy. Through the simulate analysis, indicated that theentropy combined with the wavelet multi-scaled theory can reflect the information of faultsignal in different frequency, compare to use the entropy theory alone can reflect theinformation of unsteady fault signal better. As to the feature space divided in the process ofthe entropy defined, put forward the complexity measure concept in the information theory,through analysis and compare, indicated that the feature space divided from the angle ofcomplexity can reflect the information more sensitive than the angle of the energy only. Inorder to overcome the shortcoming that the output from wavelet multi-scaled entropy is difficult to judge the fault type, introduced neural network technology as the fault typeclassifier, through training fault type neural network and simulate analysis, indicated thismethod can output the type of the gyroscope fault accurately and directly. Compare to use themethod of wavelet transform, information entropy and neural network, the algorithm thispaper introduced is combined the preparatory process, signal feature analysis and the faulttype output perfectly. Last, by simulate the gyroscope signal and analysis, find the algorithmcan detect and diagnose the fault signal effectively.