Enhanced SHM Using Tensor Decompositions and Sparse Atomistic Models for Damage Monitoring of Aeronautical Composite Structures

Abstract

Developing robust Structural Health Monitoring (SHM) solutions for large structures, particularly in the aerospace sector, remains challenging due to the volume, variability, and complexity of the data involved. One very promising solution is based on active ultrasonic guided waves (UGW) which are signals emitted and received by a set of transducers bonded to the structure to monitor. However, existing SHM algorithms cannot solve the aforementioned challenges under the current paradigm of path-by-path processing of the raw UGW signals. To move forward, a new paradigm is introduced in this work. This new approach exploits the intrinsic multi-dimensional tensorial nature of SHM UGW data through Canonical Polyadic Decomposition (CPD) and couples it with the Single Atom Convolutional Matching Pursuit Method (SACMPM). This redefines classical sparse decomposition techniques building accurate and efficient wave propagation models tailored to SHM applications. A unique UGW database where regular ground-based measurements have been carried out on an actual A380 running flight test is described in order to challenge the proposed paradigm shift. The efficiency of the coupling between SACMPM and CPD is illustrated here with respect to their ability to compress UGW information, and extract meaningful information from UGW signals in a physically informed manner. Additionally, a CPD-based damage localization algorithm is enriched using a SACMPM decomposition. Extracted physically informed features can thus be efficiently used for physically informed data driven damage monitoring approaches. The proposed paradigm shift thus demonstrates a strong potential for scalable, transferable, and reliable UGW SHM solutions, bridging the gap between laboratory experiments and real-world deployment. This paradigm shift is also expected to inspire further research and innovative ideas, leading to breakthroughs in the adoption of active UGW signals for SHM applications.