When it comes to take proper maintenance decisions regarding reliability and safety of a system, there is a need to perform a right health assessment. Thus, acquiring signals from the system in healthy and damaged conditions gives the chance to analyse the effect of the state of the system on its response. However, it is usually hard to perform diagnosis and prognosis using only tests from the real system. The advances in technologies involving internet of things, cloud computing and big data lead to a situation in which this analysis of acquired data can be complemented by the use physics-based modelling. Thus, a combination of both data-driven and physics-based approaches can be implemented thanks to the aforementioned progress. In this paper an architecture to implement hybrid modelling is proposed, based on data fusion between real data and synthetic data obtained by simulations of a physics-based model. This architecture has two analysis levels: an online process carried out in a local basis and virtual commissioning performed in the cloud. The former results in failure detection analysis for avoiding upcoming failures whereas the latter has as aim a further analysis involving both diagnosis and prognosis.