Digital Twin Modeling and Multi-Fidelity Deep Surrogate Simulation of Corrosion Fatigue Degradation Processes for the Condition-Based Maintenance of Permanent Flexible Fall Arrest Systems

Permanent flexible fall arrest systems are safety-critical equipment whose load-bearing capacity is governed by two coupled processes acting on very different time scales: a millisecond-scale transient arrest process and a multi-year corrosion fatigue degradation process. This paper develops a digital twin framework that represents both processes within a single simulation and monitoring architecture, so that the condition of an in-service system is inferred from its actual state rather than from installation design parameters. The flexible horizontal lifeline is described by a geometrically nonlinear cable formulation in which sag, pretension, and large displacement during arrest are retained, while corrosion and cyclic loading enter as time-varying parameters of the effective cross-sectional area and the residual strength. To remove the computational bottleneck of repeated high-fidelity finite element evaluation over the service horizon, a multi-fidelity deep surrogate is built from a low-fidelity analytical solver and a small set of high-fidelity simulations, with transfer learning correcting the systematic bias of the coarse model and recovering the sharp peak of the arrest force. The surrogate is then embedded in a Bayesian state updating loop driven by periodic inspection data. Across quasi-static, dynamic arrest, and long-term degradation case studies, it reproduces the peak arrest force and the anchor reaction within about 3.5% of the high-fidelity reference, reduces the evaluation time by several orders of magnitude, and tracks the residual fatigue life under realistic inspection intervals. All reference values here come from the high-fidelity model developed in this paper, with no physical test or field record, so the work is reported as a proof-of-concept framework rather than a validated engineering tool, indicating a candidate route toward condition-based maintenance planning of fall protection infrastructure.