This article proposes a new approach for the analysis of functional failure identification and propagation for Cyber-Physical Systems (CPS). Due to the increasing complexity of these systems in heterogeneous technologies, CPS require the development of more elaborate diagnostic techniques. In a first part, we propose a modelling methodology of CPS based on the notions of components and functions through four kinds of knowledge (functional, structural, topological and behavioral). In a second part, we propose an algorithm contributing to the identification of functional failures. From the definition of the CPS, a set of observation vectors is generated. This set of observation vectors is analyzed: if component failure ambiguity emerges, the set of components involved in the ambiguity of the failure is identified. Then a timed automaton is used, it performs a timing analysis of the start time and end time of each function accomplished by the supposed failed components. The timing analysis allows to remove the ambiguity in most cases. In the third part of the paper, the problem of fault propagation is studied by a mixed approach. Considering the malfunction of a component, based on the structural and topological knowledges of the CPS, a causal graph of this malfunction is generated. Then, this causal graph can easily be transformed into a fault tree (FT). A quantitative analysis of this FT can helps us to identify the cause events in the fault propagation. We illustrate our results through a sample case study.