Optimal estimation of dynamically evolving diffusivities (Q1339506)

This paper deals with the estimation and system identification of distributed systems of partial differential equations (PDEs). The augmented, iterated Kalman smoother is applied to system identification for inverse problems in evolutionary differential equations. The Kalman filter-smoother is the optimal estimator of time dependent state vectors given noisy measurements and evolution equations with stochastic forcing. The Kalman smoother minimizes a quadratic functional which includes the residual squared error to the measurements. In the augmented smoother, the unknown, time dependent coefficients are included in the state vector and have a stochastic component. At each step in the iteration, the estimate of the time evolution of the coefficients is linear. The present approach is common for finite dimensional engineering problems, but it is unaware of any previous work which uses the augmented Kalman smoother to estimate unknown coefficients in distributed systems of partial differential equations. The Kalman filter-smoother with smoothness penalty functions and its implementation features are briefly and clearly described in the five appendices of the paper. Potential applications are described which include the estimation of anomalous diffusion coefficients in turbulent fluids and the plasma rotation velocity in plasma tomography.