Digital twinUS & West
Streaming Sparse Cholesky for Derivative-Informed Gaussian Process Surrogates in Digital Twins
The paper presents an end-to-end digital twin solution for predicting the state of an aircraft structure. The method extends Gaussian processes to include derivative data for improved accuracy and uses a streaming sparse Cholesky factorization for efficient updating with in-service data. This enables real-time high-fidelity forecasting.