A Dynamic System Structure Modeling and Evolution Identification Method Based on Multi-Source Signal Fusion

Dynamic systems often evolve over time under complex environments, where structural changes and nonlinear behaviors can be induced by multi-source signals. This paper proposes FUSE-ID (Fusion-based Structure Evolution and Identification), a novel framework for modeling and identifying the dynamic structural evolution of systems through multi-source signal fusion. Unlike traditional single-source approaches, FUSE-ID integrates heterogeneous signals—such as sensor data, time series, and latent states—using an adaptive fusion module that captures inter-source dependencies and contextual dynamics. At the core of FUSE-ID is a dynamic structure modeling mechanism that adaptively captures evolving topologies, enabling fine-grained representation of system changes over time. Complementing this, a self-supervised evolution identification module is embedded, enabling the accurate detection of structural shifts and system anomalies. Experiments on synthetic and real-world datasets demonstrate that FUSE-ID outperforms existing approaches in terms of modeling accuracy, evolution tracking, and robustness to noisy observations. This work provides a promising avenue for intelligent monitoring, fault diagnosis, and adaptive control in complex dynamic environments.