Digital Twin Behavior Matching of Gas Plumes Using a Fixed Sensor Mesh and Dynamic Mode Decomposition

Digital twins (DT) have become a useful tool in smart manufacturing, engineering and controls. Behavior matching of DTs to their physical twin counterparts is essential for capturing the evolution of key system parameters. Given that environmental gas emissions are governed by partial differential equations, the behavior matching optimization can often be ill posed and computationally expensive. Stochastic models have shown good agreement to deterministic models while having a significant computational cost reduction. This work presents a method for solving the source localization problem using a DT implementation of a stochastic point source emission with a fixed-mesh of gas sensors. The DT source localization is determined through behavior matching process with low frequency modes after dynamic mode decomposition using spatial interpolation on measured time series data. That is, the minimization of the mismatch between the DT and the unknown physical model can given an estimate of the source location.