Abstract
Accurately estimating spatially heterogeneous elasticity parameters,particularly Young's modulus and Poisson's ratio, from noisy displacementmeasurements remains significantly challenging in inverse elasticity problems.Existing inverse estimation techniques are often limited by instability,pronounced sensitivity to measurement noise, and difficulty in recoveringabsolute-scale Young's modulus. This work presents a novel Inverse ElasticityPhysics-Informed Neural Network (IE-PINN) specifically designed to robustlyreconstruct heterogeneous distributions of elasticity parameters from noisydisplacement data based on linear elasticity physics. IE-PINN integrates threedistinct neural network architectures dedicated to separately modelingdisplacement fields, strain fields, and elasticity distributions, therebysignificantly enhancing stability and accuracy against measurement noise.Additionally, a two-phase estimation strategy is introduced: the first phaserecovers relative spatial distributions of Young's modulus and Poisson's ratio,and the second phase calibrates the absolute scale of Young's modulus usingimposed loading boundary conditions. Additional methodological innovations,including positional encoding, sine activation functions, and a sequentialpretraining protocol, further enhance the model's performance and robustness.Extensive numerical experiments demonstrate that IE-PINN effectively overcomescritical limitations encountered by existing methods, delivering accurateabsolute-scale elasticity estimations even under severe noise conditions. Thisadvancement holds substantial potential for clinical imaging diagnostics andmechanical characterization, where measurements typically encounter substantialnoise.