Satchi Venkataraman, Ph.D., is a Professor of Aerospace Engineering at San Diego State University. His expertise and interests are in the Analysis and Design of Aerospace Structures, Structural Optimization, Uncertainty Quantification, and Risk Assessment. His current research investigates models for failure prediction in aerospace composite structures; understanding effects of defects in composite materials and structures, design for damage tolerance and predictable failure, use of surrogate models for multi-level optimization, and biomimetic design of structures.
Bearing failure prediction of bolted joints in composites is complicated due to the interactions of contact, geometric, and material nonlinearity, as well as multimodal and multiscale nature of damage and failure of CFRP materials. Hybrid composite fiber metal laminates (FML), that include metal foils along with carbon fiber reinforced polymer (CFRP) plies, exhibit increased laminate bearing strength. In FML, the ductility of the added metal layers further complicates this, requiring changes to models previously developed for bearing failure of CFRP bolted joints. High-fidelity prediction models that use finite element methods are computationally expensive and prohibitive for use in design. Low-fidelity models make simplifications in the representation of geometry, loading, and material models used for failure and produce computationally efficient models with lower accuracy. These low fidelity models, prevalent in design, require empirical correction factors and are difficult to generalize for arbitrary cases. Combining models of varying fidelities to develop multi-fidelity analysis tools requires understanding how model fidelity affects prediction accuracy and computational cost. Despite the vast array of modeling and analysis choices developed for composite bolted joints, there is no established method to classify and compare or rank models of different fidelity and correlate them to their prediction accuracy and computational costs. This talk presents a framework to quantify or rank modelling choices based on the fidelity in representation of geometry, load, composite laminate representation, finite elements used and meshing, and constitutive models for the damage and failure. The exploration of modeling choices for bearing failure prediction in bolted joints in fiber metal laminates and their effects on affects the prediction accuracy and computational cost are discussed. Challenges in model validation of high-fidelity models is highlighted. The framework developed will enable the development of physics informed machine learning multi-fidelity models for predicting bearing failure of bolted joints in composite structures.