The Improvement Of Conventional Principal Component Analysis And Research On Online Fault Diagnosis

Nowadays, Autonomous Underwater Vehicles (AUVs) are widely used in commercial, scientific, military and many other fields for various purposes. Autonomous Underwater Vehicles, as an important component of high marine technology, requires higher intelligent level. It is extremely necessary to research underwater robot fault diagnosis technique, thus robot can safely execute and complete underwater tasks in complicated underwater environment. Fault diagnosis technique also embodies underwater robot's intelligence.Sensors are an important component to detect Underwater Vehicles. Effective fault detection and diagnosis to sensors can not be ignored for the safety of underwater robots. When robots are operating underwater, the sensor fault will cause obstacles to the successful completion of operation and may even lead to disastrous consequences and non-valuation losses.In order to study the feasiblity and effectiity of underwater vehicles' online fault diagnosis methods based on Wavelet Denoising and PCA fault diagnosis, this paper introduce wavelet theory and principal component analysis to the underwater Vehicles' fault diagnosis.During underwater missions, the sensor will be interfered by the flow noise, etc. When diagnosing an underwater robot sensor with conventional principal component analysing fault diagnosis method for fault diagnosis,the principal component model and the sensor itself both contain sensor noise signal,so we can not guarantee the accuracy of the principal component model or diagnose sensor fault is affected by noise or sensor failure in itself. Introduce the signal pre-processing process of wavelet denoising to the sensor signals based on Wavelet Denoising and PCA fault diagnosis method, avoid noise signals impacting on the diagnosis results and emerging misjudgment phenomenon, guarantee the accuracy of fault diagnosis results and make fault diagnosis results truly reflect the sensor fault condition. verify the feasibility of wavelet denoising and PCA fault diagnosis method with the historical data of some kind of underwater robot.Propose an online fault diagnosis method for online fault diagnosis problems of underwater robots. Analyze principal component modeling problems of underwater vehicles' online fault diagnosis based on Wavelet Denoising and PCA fault diagnosis method and propose a set of principal component modeling methods combining historical data models and the advantages of the scene modeling. The test results of underwater vehicle "Beaver" has verified the feasibility of this methods.