Damage detection at an aluminum beam from discrete and continuous strain measurements

Unless the sensors are closely located to a local defect, the change in the global strain field caused by the defect is very small, and may go faded by other environmental effects. Only when compared the strain readings at many points, some information about damage may be unveiled. Robust automated techniques are needed to do this comparison. Principal Component Analysis (PCA) is a well-known statistical technique that has been used as a pattern recognition technique by several years with excellent results. It allows obtaining pattern that often underlie from the data by calculating the principal components and re-expressing the information in a new space. Damage index are already available. An experimental validation of the technique is discussed in this paper, comparing damages of different sizes and positions, under a set of combined loads, both under static and dynamic conditions. Strains were measured at several points by bonded FBGs (Fiber Bragg Gratings), and also along continuous lines by optical fiber distributed sensing (OBR, Optical Backscatter Reflectometer). The sensitivity of the approach and the influence of parameters (number of sensors, distance to the damage) are quantified.