Comparison of two robust PCA methods for damage detection in presence of outliers

Statistical methods such as Principal Component Analysis (PCA) are suffering from contaminated data. For instance, variance and covariance as vital parts of PCA method are sensitive to anomalous observation called outliers. Outliers, who are usually, appear due to experimental errors, are observations that lie at a considerable distance from the bulk of the observations. An effective way to deal with this problem is to apply a robust, i.e. not sensitive to outliers, variant of PCA. In this work, two robust PCA methods are used instead of classical PCA in order to construct a model using data in presence of outliers to detect and distinguish damages in structures. The comparisons of the results shows that, the use of the mentioned indexes based on the robust models, distinguish the damages much better than using classical one, and even in many cases allows the detection where classic PCA is not able to discern between damaged and non-damaged structure. In addition, two robust methods are compared with each other and their features are discussed. This work involves experiments with an aircraft turbine blade using piezoelectric transducers as sensors and actuators and simulated damages.