Non-linear Damage Classification based on Machine Learning and Damage Indices

The use of guided wave-based approaches presents some advantages in the structural inspection and damage identification processes. It is driven by the fact that these waves can propagate over relatively long distances and are able to interact sensitively with and uniquely with different types of defects, however, its use in Structural Health Monitoring requires the development of efficient SHM methodologies to analyse and provide confident results. To do that, signal processing techniques for the correct interpretation of the complex ultrasonic waves are a need. In this sense, it is necessary to still work on the continuous search of methodologies for performing each one of the steps in the damage identification. As contribution, this paper presents a damage classification methodology which includes the use of data collected from a structure under different structural states by means of a piezoelectric sensor network. The document presents the description of the methodology including a description of the data reduction and the use of non-linear analysis of the information with hierarchical non-linear principal component analysis and some non-linear damage indices. The methodology is preliminary evaluated with a CFRP sandwich structure with some damages on the multi-layered composite sandwich structure which were intentionally produced to simulate different damage mechanisms, i.e. delamination and cracking of the skin. Finally, results are presented and discussed to remark the advantages and disadvantages of this methodology.