A fast algorithm to racetrack filter multiaxial histories preserving load shape

Abstract

A fast algorithm is presented to implement a multiaxial version of the racetrack amplitude filter, improved upon the authors' previous works. The algorithm only requires a single user-defined scalar filter amplitude to eliminate noise and potentially non-damaging events from arbitrary non-proportional multiaxial stress or strain histories, while preserving all key features of the load path. It can be applied as a preprocessing step to both invariant-based and critical-plane multiaxial damage models, highly decreasing the computational cost in assessments involving oversampled or noisy data, much improving the usefulness of computationally-intensive multiaxial models. Since load path shape is well preserved (within the chosen filter resolution), peaks and valleys from subsequent candidate plane projections are not filtered out. Moreover, path-equivalent stress or strain calculations can still be applied after the filtering process, without precision loss beyond the filter resolution. The algorithm is synchronous and preserves load order, becoming suitable for strain-based multiaxial damage models involving plastic memory, as well as for mixed-mode crack propagation calculations. The filter efficiency is validated from oversampled tension-torsion experimental data in 316L stainless steel tubular specimens, following complex non-proportional histories. Finally, a fully-tested computer implementation of the algorithm is presented.