AI-enabled high-throughput characterization of material properties based on indentation response

High-throughput characterization (HTC) is a highly time-efficient and spatially compact approach for materials testing. It enables acquisition of comprehensive material properties in a single experiment. Among the associated testing techniques, indentation-based materials characterization is of significant importance. Given the mappings of modulus distribution on indentation responses derived by symplectic contact analysis, as well as finite element simulations, a physics-informed U-net enhanced Fourier Kolmogorov–Arnold neural operator (PIU-FKANO) is established as a surrogate model to carry out inverse contact analysis. Displacement and curve measurement techniques are applied to facilitate integration of the neural operator with experiments. Predicted and true modulus fields demonstrate a high consistency in four representative numerical examples. Different architectures of PIU-FKANO are analyzed to explore a model that excels in terms of the relative L2 error. Furthermore, the proposed model also shows more stable performance than other models, thus offering robust support for HTC techniques.