Real-time thermoacoustic data assimilation

Low-order thermoacoustic models are qualitatively correct, but typically, they quantitatively inaccurate. We propose a time-domain bias-aware method to make low-order (more) accurate. First, we develop Bayesian ensemble data assimilation for model self-adapt and self-correct any time that reference become available. Second, apply the methodology infer states heat-release parameters on fly without storing (real time). perform twin experiments using synthetic acoustic pressure measurements analyse performance of in all nonlinear regimes, from limit cycles chaos, interpret results physically. Third, practical rules assimilation. An increase, reject, inflate strategy is proposed deal with rich behaviour; physical scales non-chaotic regimes (with Nyquist–Shannon criterion) chaotic Lyapunov Fourth, higher-fidelity model. introduce an echo state network estimate real forecast bias, which error low-fidelity show that: (i) correct pressure, parameters, bias can be inferred accurately; (ii) learning robust as it tackle large uncertainties observations (up 50 % mean values); (iii) uncertainty prediction naturally part output; (iv) both time-accurate solution statistics successfully. Data opens up new possibility real-time thermoacoustics by combining knowledge experimental synergistically.