Data-driven prognostics for fiber reinforced composites based on multimodal NDE monitoring

Multimodal Nondestructive Evaluation (NDE) is used in this investigation to estimate the evolving material state in carbon fiber reinforced composite specimens to subjected mechanical loading. Specifically, Acoustic Emission (AE) monitoring, Digital Image Correlation (DIC) measurements and passive Infrared Thermography (pIRT) are used to monitor damage in IM7-8552 specimens. Both pristine Straight Edge (SE) and Open Hole (OH) specimens were loaded until failure to monitor the damage evolution and link this to the observed NDE trends. Post mortem microscopy coupled with in situ NDE identified that while multiple damage mechanisms were present, delamination is the dominate mode of catastrophic failure. The in situ monitored NDE data trends are leveraged in a machine learning framework combining unsupervised clustering and outlier analysis to denoise the recorded data and produce a degradation curve that can be used in a diagnostics and prognostics FEA approach. A computational framework capable of performing 3D FEA analysis and providing predictions of the distributed damage within the composite layers using real-time recorded damage data trends is presented and the use of this methodology is discussed in terms of its material and loading agnostic nature and in particular its applicability and potential in forming data-driven prognostics of remaining useful life.