1: \begin{abstract}

2: This paper deals with the observer design problem for time-varying linear infinite-dimensional systems.

3: We address both the problem of online estimation of the state of the system from the output via an asymptotic observer,

4: and the problem of offline estimation of the initial state via a Back and Forth Nudging (BFN) algorithm.

5: In both contexts, we show under a weak detectability assumption that a Luenberger-like observer may reconstruct the so-called observable subspace of the system.

6: However, since no exact observability hypothesis is required, only a weak convergence of the observer holds in general. Additional conditions on the system are required to show the strong convergence.

7: We provide an application of our results to a batch crystallization process modeled by a one-dimensional transport equation with periodic boundary conditions, in which we try to estimate the Crystal Size Distribution from the Chord Length Distribution.

8: \end{abstract}

9: