The most important portion of an expert system development is the articulation of knowledge by the expert and its satisfactory formulation in a suitable knowledge representation scheme for mechanization by a computer. A "deep knowledge" approach called Goal Tree-Success Tree model is devised to represent complex dynamic domain knowledge. This approach can hierarchically model the underlying principles of a given process domain (for example nuclear power plant operations domain). The Goal Tree-Success Tree can then be used to represent the knowledge-base and provide means of selecting an efficient search routine in the inference engine of an expert system.