Antonio Bilotta

Determining the band gaps of metamaterials based on well-defined unit cells is a crucial line of investigation, as it provides fundamental design principles for tailored mechanical metamaterials with targeted properties. Practical applications notably include frequency-filtering devices, which are especially advantageous for structural vibration mitigation and acoustic insulation. For elementary unit cell geometries, dispersion relations can be derived analytically, yielding direct estimates of band-gap boundaries and actionable insights into parameter tuning. However, as unit-cell topologies become increasingly intricate -- a distinctive feature of modern metamaterials -- analytical approaches become intractable, necessitating numerical computational strategies. In this study, we investigate the dynamic behaviour of planar pantographic beams using a discrete element model composed exclusively of translational and rotational springs. Through numerical experiments, we delineate the parameter space governing band-gap formation, providing concise, practical guidelines for optimal unit-cell design. Ultimately, the framework presented herein offers a generalizable methodology for analysing more complex and topologically distinct unit-cell configurations.

Is it possible to interpret the modeling decisions made by a neural network trained to simulate the constitutive behavior of simple or complex materials? The problem of the interpretability of a neural network is a crucial aspect that has been studied since the first appearance of this type of modeling tool and it is certainly not specific to applications related to constitutive modeling of heterogeneous materials. All areas of application, such as computer vision, biomedicine, and speech, suffer from this fuzziness, and for this reason, neural networks are often referred to as "black-box models". The present work highlighted the efforts dedicated to this aspect in the constitutive modeling of the behavior of path independent materials, reviewing both more standard neural networks and those adopting, more or less strongly, the specific point of view of interpretability.

The identification of damage in the connection of steel–concrete composite beams is pursued by means of Convolutional Neural Networks. The data used for training the networks are gray level images obtained by converting a set of transmissibility functions of the beam. It is shown how simple Convolutional Neural Networks can be trained for the identification of the damage by evaluating its position and intensity. Data availability is thoroughly discussed highlighting the effects of various factors, among them the number and richness of the images used in the training phase, on the predictive capabilities of the network. In this regard, the results obtained are satisfactory as confirmed also by comparisons with data extracted from an experimental campaign. The problem of the unavoidable presence of modelling errors is also discussed and counteracted.

A symbolic mathematical approach for the rapid early phase developing of finite elements is proposed. The algebraic manipulator adopted is MATLAB® and the applicative context is the analysis of hyperelastic solids or structures under the hypothesis of finite deformation kinematics. The work has been finalized through the production, in an object-oriented programming style, of three MATLAB® classes implementing a truss element, a tetrahedral element and plane element. The approach proposed, starting from the mathematical formulation and finishing with the code implementation, is described and its effectiveness, in terms of minimization of the gap between the theoretical formulation and its actual implementation, is highlighted.