Invisible Hydrodynamic Sensing via Metamaterial Shells Optimized by Machine Learning

Accurate flow-field sensing is crucial for microfluidics, biomedicine, and environmental monitoring. However, conventional sensing devices inherently distort the flow due to permeability mismatch, leading to erroneous data. To overcome this fundamental limitation, we propose and validate a hydrodynamic sensing mechanism based on a rationally designed metamaterial shell that creates a protected sensing core while leaving the external flow almost undisturbed. Specifically, the shell restores the pressure field in the core region to the background pressure field that would exist in the absence of any obstacle, so that an actual sensor placed there would read the true undisturbed pressure, and at the same time suppresses the disturbance of the combined core-shell structure to the surrounding flow. This shell, with anisotropic permeability derived from scattering-cancellation theory, thus renders the sensing core invisible to the background flow. Navigating the vast design space is achieved via a deep neural network, which inversely designs microstructures with exceptional accuracy (prediction error <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:semantics><mml:mrow><mml:mo><</mml:mo> <mml:mn>1</mml:mn></mml:mrow> <mml:annotation>$< 1$</mml:annotation></mml:semantics> </mml:math> %). Our metashell-enabled sensing scheme reduces pressure measurement errors by four to five orders of magnitude, achieving near-perfect fidelity even under severe permeability contrasts. This theoretical-machine learning framework establishes a blueprint for distortion-free hydrodynamic sensing, readily extendable to thermotics, acoustics, and electromagnetics.