Damage detection using robust PCA

Contam nated data out ers can affect the resu ts of stat st ca methods such as Pr nc pa Component Ana ys s An out y ng observat on or out er s one that appears to dev ate marked y from other members of the samp e n wh ch t occurs Out ers usua y appear due to exper menta errors and cou d have undesr ab e nf uence on the resu t of the method An effect ve way to dea w th th s prob em s to app y a robust e not sens t ve to out ers var ant of PCA In th s rev ew d fferent robust PCA methods are summer zed br ef y; then the most two mportant robust PCA methods are used nstead of c ass ca PCA The resu ts from both methodo og es robust and c ass c are compared The compar sons of the resu ts shows that the use of the ment oned ndexes based on the robust mode s d st ngu sh the damages much better than us ng c ass ca one and even n many cases a ows the detect on where c ass c PCA s not ab e to d scern between damaged and non- damaged structure As we as compar son between c ass ca and robust methods two robust methods are a so compared w th each other and the r features are d scussed Requ red data are co ected us ng an a rcraft turb ne b ade ut z ng p ezoe ectr c transducers as sensors and actuators and d fferent s mu ated damages Structural health monitoring (SHM) is a damage detection technique that involves placing intelligent sensors on a structure, periodically recording data from the sensors, and using statistical methods to analyze the data in order to assess the condition of the structure. The field of SHM developed through the combination of nondestructive evaluation (NDE) methods and novel sensing and actuation techniques to create intelligent monitoring systems permanently installed on structures, in other words, SHM is a non�괂Ġ destructive method aims to give, at every moment during the life of a structure, a diagnosis of the "state" of the constituent materials, of the different parts, and of the full assembly of these parts constituting the structure as a whole. It involves the integration of sensors, possibly smart materials, data transmission, computational power, and processing ability inside the structures. It makes it possible to reconsider the design of the structure and the full management of the structure itself and of the structure considered as a part of wider systems. Structural health monitoring offers a powerful method of monitoring structures with a promise of shifting time�괂Ġ based