Multi-level damage diagnosis on stiffened composite panels based on a damage-uninformative digital twin

Abstract

In this study, a multi-level Structural Health Monitoring methodology for stiffened composite panels is introduced. A digital twin (DT), that is, a three-dimensional finite element (FE) model, representing the pristine state baseline of the test article, is developed and verified for compressive loading in the post-buckling regime. The detailed FE model is utilized to train a surrogate model with respect to exogenous input, that is, axial load magnitude. The surrogate assists the DT concept that would allow prediction of the load acting on the structure based on an influx of strain data, acquired from fiber Bragg grating sensors permanently attached along the stringer feet. For this purpose, we leverage on the observation that remote from the damage, the strain field remains virtually unaltered with regard to the pristine state. The load is estimated by a sensor placed far from the damage whilst the diagnostic actions are performed by exploiting measurements from the remaining sensing locations. A health indicator, which compares the experimentally received strains with those from the surrogate representing the pristine state, is utilized to (1) detect, (2) localize, and (3) characterize the damage. As damage, we consider either skin-to-stringer disbond or initial impact damage propagation as well as overall stiffness degradation during thousands or millions of fatigue cycles. The sensors that have detected a disbond are dedicated to evaluating the potential propagation of it, while the remaining sensors evaluate the overall stiffness degradation. The proposed methodology is tested for one artificially disbonded and two impacted single-stringer panels subjected to block loading compression-compression fatigue.

Digital Twin-Based Damage Quantification on Composite Structures

“…The offline stage will constitute the digital counterpart of the physical stiffened panel which will be feeding strains, during the online stage, to the optimization scheme in order to quantify the disbond area and location. The methodology also takes into account the variability on the compressive load applied to the panel by performing load identification prior to the disbond quantification [20,21].…”

“…In the final step, the inverse quantification of the disbond parameters are described. In our methodology, we leverage on the notion that far from damage the strain field remains unaltered [21]; hence, strain readings from sensor R1-R4 are used to infer the load applied on the panel, based on the pristine numerical model. The steps of the damage quantification are presented below:…”

“…Select reference sensor(s) (R1,..,R4) for load prediction 2. Infer applied load via gradient-descent optimization [21] 3. Infer the deterministic disbond parameters via minimizing the following objective function for a measurement point, m:…”

Graphene nanoparticles as data generating digital materials in industry 4.0

“…Finally, in the operational phase, DT is used for prognostics and diagnostics activities 108 . Milanoski et al 109,110 www.nature.com/scientificreports/ of stiffened composite panels by estimating the load acting on the structure using strain data acquired from Fiber Bragg Grating (FBG) sensors. Sisson et al 111 pursued a digital twin approach to optimize rotorcraft flight parameters by minimizing stress on critical mechanical components and through probabilistic diagnosis, prognosis, and optimization.…”

“…Sisson et al 111 pursued a digital twin approach to optimize rotorcraft flight parameters by minimizing stress on critical mechanical components and through probabilistic diagnosis, prognosis, and optimization. Using the data collected from strain sensors, it is not only possible to detect the presence of the damage but also the evolution of the damage, hence remaining useful life of the part can also be predicted 109 .…”