In this paper a structured knowledge-based approach to the representation and scheduling of flexible manufactoring systems (FMSs) is described. Our approach is based on a structured conceptual representation (a KL-ONE-like Si-net representation formalism), extended with an instant-based temporal reasoning formalism. Furthermore, the approach integrates a particular extension to high-level Petri nets (PNs), structured timed colored Petri nets (STCPNs), for the modeling and simulation of the FMS. Such a representation scheme allows us to use SI-nets' good properties related to inference (classification and inheritance), which are lacking in PNs, and at the same time provides an extension toward an explicit representation for time. The integration of Si-nets with PNs is necessary because simulation and low-level coordination of FMSs require a procedural approach that is not within the aims of Si-nets. Therefore, procedural and symbolic levels, corresponding to the different hierarchical levels of the representation and control system of the FMS, coexist in the system. Using a qualitative terminology, we may also call them analog and symbolic knowledge. We assume that such a hybrid representation system may be useful, since a procedural representation, integrated within a logic formalism, can increase the expressive power without complicating the notation or the representation itself. The paper describes both the representational aspects and the modeling of the control system of the FMS, focusing on the interaction mechanisms among the different levels of representation. In particular, we show how an STCPN-based model can be automatically derived starting from the symbolic component of the system. A particular FMS case study, regarding a class of problems of resource-constrained multiproject scheduling (where projects are sets of tasks temporally related), is discussed.