Domain-Shift Aware Neural Networks for Unbalance Characterization in Rotating Systems

Researchers have developed new neural network frameworks for solving complex systems, including domain-shift aware models for estimating unbalance masses in rotating shafts and randomized networks for Poisson-Nernst-Planck and Poisson-Nernst-Planck-Navier-Stokes systems.

A team of researchers has made advancements in applying neural networks to complex problems. One study used a domain-shift aware neural network to estimate unbalance masses in rotating shafts under varying conditions^[1]. The network was trained with a maximum mean discrepancy strategy to align feature representations across different distributions. Experimental data were collected from a test rig with a primary and secondary shaft, using triaxial accelerometers to record the dynamic response. The results showed improved prediction accuracy by addressing domain shift. In a separate study, a new framework called SO-RaNN was developed for solving Poisson-Nernst-Planck and Poisson-Nernst-Planck-Navier-Stokes systems using randomized neural networks^[2]. The framework solves decoupled linearized subproblems iteratively and employs a pointwise cut-off to enforce positivity. An SAV-type post-processing correction was also used to introduce an auxiliary discrete dissipation mechanism.