Real time fatigue detection in structures by means of strain field pattern recognition

Usually fatigue cracks are detected by means of nondestructive testing, which involves long examination times and qualified personal. The detection process is then, expensive and time consuming. Recent advances in Structural Health Monitoring (SHM) techniques are very promising for the damage assessment in complex structures; however, fatigue detection is still one of the main challenges and open topics in the research field. Few techniques has proven effective for real time fatigue detection during the operation of structures. This paper presents a novel methodology for fatigue detection in structure under operational conditions. The methodology is based on strain measurements and strain field pattern recognition; in turn, pattern recognition is based on dimensional reduction and feature extraction techniques using neural networks. The advantages of the Fiber Optic Sensors (FOS) for strain measurements are exploited, in particular, Fiber Bragg Gratings (FBGs). Unless the strain sensors are closely located to a local fatigue crack, the change in the strain field caused by the crack is very small. Only when strain readings at several points are compared (pattern recognition), some information about fatigue damage may be unveiled. Robust automated techniques are needed to perform such comparison taking into account the redundant or useless information. Two different techniques are proposed for feature extraction: Principal Component Analysis (PCA) and Hierarchical Nonlinear PCA. An experimental validation of the technique is discussed in this paper.