Mr. Guy Priebatch - M.Sc. Candidate
03/09/2026
אודיטוריום ע"ש דויד וואנג, בניין מידן, קומה 3
13:30
Engineering structures routinely experience a combination of forces and loading modes throughout their life cycles, which may include tension, torsion, and bending. In standard bulk materials, the stiffness and overall response across these varied loading modes are inherently coupled. This makes programming each mechanical mode separately almost impossible. In this talk, I present a framework that uses helical fiber-embedded cylindrical composites, with tunable volume fraction, pitch, and helical radius, to decouple and program mode-specific mechanical responses. This leads to an inverse design challenge, in which the helical fiber geometry parameters are computed to achieve prescribed target behavior. To address this, I employ an artificial neural network (ANN) as an inverse solver. To train the network, load-deformation data is generated from finite-element simulations and mechanical characterization and testing of 3D-printed specimens. These load-deformation curves are fitted to a polynomial series to capture the non-linear behavior of the composites. The trained ANN accepts the series coefficients as inputs and predicts the corresponding helical-fiber geometric parameters. The framework is applied to design individual composite rods and structures that achieve target responses under different loading conditions.
The proposed approach breaks the classical stiffness trade-offs through geometry rather than composition: microstructural design alone can significantly expand the range of accessible mechanical properties from a limited set of base materials. This opens a practical path to mechanical performance that can be tailored on demand for different applications.