High-dimensional data analytics applications in SHM and NDE: tensor analysis of thermal videos

Abstract

This paper discusses the established and potential applications of high-dimensional data analysis in the fields of structural health monitoring and non-destructive evaluation. Despite the significant potential of high dimensional data analytic methods, a few applications have been implemented in structural health monitoring and non-destructive evaluation. Further, as measuring technologies improve, the requirement of applying these approaches grows. This paper uses thermal videos as an example of high-dimensional data in the non-destructive evaluation field. These thermal videos are used to detect and localize delamination in composite plates, typically found in aircraft wings. Using traditional statistical approaches to analyze videos presents theoretical and practical challenges due to their high dimensionality. Tensor analysis methods help to overcome these issues. To locate the damage, two tensor factorization algorithms are used. For a rectangular damage zone, two vectors are enough to localize the extent of the damage. For more sophisticated cases like a damage in the shape of a circle, higher order of core tensors with larger projection tensors are needed. The results demonstrate these methods are accurate and efficient in terms of computing cost.