Jonathan Hopkins1
University of California, Los Angeles1
Jonathan Hopkins1
University of California, Los Angeles1
The objective of this work is to apply the concept of artificial neural networks (ANNs) to enable the creation of a new kind of architected material, called a mechanical neural-network (MNN) architected material, that can learn desired properties via a complex web of active flexible elements (AFEs) that constitute the materials’ microstructure. These AFEs, which are joined together by rigid nodes, constitute an analogous physical embodiment of the mathematical weights that determine the values that are summed together by the neurons within traditional ANNs. By actively tuning the stiffness of these AFEs in a similar fashion to how weights are trained within ANNs, the new kind of architected material can learn desired mechanical properties and thus enable a variety of applications. Such applications include: (i) aircraft wings that can learn to optimally change their shape and stiffness at select locations as flight conditions change, (ii) aircraft exteriors that can learn to compensate for damage that may occur during combat or overuse by maintaining their designed properties regardless of defects or ware, and (iii) electrical, optical, or other components within aircraft that cannot be made of zero-thermal-expansion-coefficient materials but can learn to maintain their shape regardless of fluctuating temperatures.