Data-Driven Physics-Based Digital Twin for Linkage Analysis

Linkage mechanisms are fundamental in the design of deployable space structures, enabling efficient packing and significant expansion. However, they have a high risk of jamming and are a source of single-point failures. Simulation tools provide a viable method to increase confidence of mission success, offering insights into system behavior without relying on physical prototypes. However, rigid-body dynamic simulations struggle to capture the effects of joint clearance and friction. This paper implements a data-driven, physics-based digital-twin framework for linkage analysis. Our approach uses a differentiable physics engine and real data from hardware experiments to estimate joint clearance and friction for a target mechanism. We validate our approach using both synthetic data and a hardware experiment on a multi-cell scissor mechanism. Our model correctly captures linkage jamming due to joint clearance and friction and closely matches trajectories captured during hardware experiments.