A Two-Scale Multi-Physics Deep Learning Model for Smart MEMS Sensors

Smart materials and structures, especially those bio-inspired, are often characterized by a hierarchy of length- time-scales. Micro Electro-Mechanical Systems (MEMS) also different physical phenomena affecting their properties at scales. Data-driven formulations can then be helpful to deal with the complexity multi-physics governing response external stimuli, optimize performances. As an example, Lorentz force micro-magnetometers working principle rests on interaction magnetic field current flowing inside semiconducting, micro-structured medium. If alternating properly set frequency is let flow longitudinally in slender beam, system driven into resonance sensitivity may result largely enhanced. In our former activity, reduced-order model movable structure single-axis MEMS magnetometer was developed, feed multi-objective topology optimization procedure. That model-based approach did not account for stochastic effects, which lead scattering experimental data micrometric length-scale. The formulation here improved allow effects through two-scale deep learning designed as follows: material scale, neural network adopted learn mechanical polysilicon induced its polycrystalline morphology; device further most important geometric features parts that affect overall performance magnetometer. Some preliminary results discussed, extension size finally foreseen.