Model-driven software development considers models as a core artefact for generation of software source code. This requires models to be formal and complete enough for further transformations and code generation. It requires clear understanding of such knowledge as functionality, objects and dependencies in the problem domain. In our approach, this knowledge is kept in the frame-based system. The completeness and consistency of the knowledge can be verified by generating and validating the topological functioning model (TFM). The TFM is a model, which elements are linked by the topology, i.e. by cause and effect relations among the functional characteristics of the domain. Automated composition of the TFM requires retrieving appropriate conditions on cause and effect functional characteristics of the system from the knowledge base. The proposed algorithm reads data of functional characteristics kept in the knowledge base, relates those of them, where a cause condition corresponds to an effect condition, and generates data for the corresponding cause-and-effect relation. The difficulty is that conditions can be combined using logical operators AND, OR, XOR, as well as can use negation NOT. The benefit is that any inconsistency in the retrieved topology could be discovered and marked for further analysis. This should force careful analysis of the problem domain before generation of the design model. That could lead to decreasing a number of errors made due to uncertainty in the analysis.