Digital twins (DTs) usually consist of several parts, often described by linear time-invariant (LTI) systems, that is, systems with inputs and outputs (I/O systems) that realize the interoperation of the different submodels. Examples include mathematical models for observers and controllers, or discretized models of partial differential equations, modeling, e.g., thermodynamic or mechanical processes, etc. High-fidelity models of the physical processes comprising the physical twin are often too large to allow real-time operation of DTs. Therefore, often fast-to-simulate surrogate models are employed. These models are approximations of the high-fidelity models of the underlying physical processes. It is therefore of great importance that the interplay of the surrogate submodels correctly reflects the physical properties of the original interconnected system. In other words, the interconnection of the surrogate models should be robust to perturbations, like those introduced by the approximation errors of the surrogate. In this talk, we will focus on the interconnection of stable models. It is well known that their interconnection may not be stable, even if the individual coupled submodels are all stable, while the interconnection of passive submodels will be passive again [2]. In recent years, modeling passive systems in the port-Hamiltonian (pH) framework has gained significant interest as an energy-based modeling framework. Due to the aforementioned result, the interconnection of pH systems is pH again. Thus, the dissipativity property of a DT, or one of its parts, is guaranteed if the used submodels are written in pH form. Here, we will in particular consider the question of optimal robustness of the dissipativity property against perturbations in order to maximize the reliability of a DT consisting of dissipative submodels when approximation errors as well as unmodeled dynamics need to be taken into account. For this purpose, we will employ a certain freedom in the parametrization of dissipative models [1]. REFERENCES [1] Benner P., Van Dooren P., Goyal P.K., Parameterized interpolation of passive systems, SIAM Journal on Matrix Analysis and Applications, Vol. 45, Issue 2, pp. 1035–1053, 2024. [2] Brune O., Synthesis of a finite two-terminal network whose driving-point impedance is a prescribed function of frequency, Doctoral thesis, MIT, 1931.