A strain-interfaced digital twin solution for corner fatigue crack growth using Bayesian inference

Abstract

This paper introduces a digital twin solution for corner fatigue crack growth assessment. The digital twin comprises three core features: (1) diagnosis, (2) prognosis and (3) updating. The diagnosis arm performs remote crack size measurement via strain data collected from strategically identified locations. The prognosis component postulates the fatigue life across both linear-elastic and elasto-plastic loading regimes through a fatigue crack growth power law with the cyclic J-integral, ΔJ, as the crack driving force. Uncertainty in power law parameters, however, may result in differences between the prognosis and observed fatigue life. Hence, the digital twin completes the feedback loop via Bayesian updating of the power law parameters, thereby mirroring its physical counterpart closely. An improved estimation of the remaining useful life follows. The proposed digital twin solution validates against three specimens under constant amplitude loading and a single specimen under variable amplitude loading. The successful application of the approach marks a significant step toward operational digital twins within practical settings.