The conventional Petri net approach to verify knowledge bases was found not to be adequate to verify the knowledge bases of large and complex systems such as alarm processing systems of nuclear plants. Jensen has proposed Colored Petri Net that gives colors to the token of Petri Net to model large systems with small nets. However, Colored Petri Net also still has some limitations in modeling the knowledge base. We propose an improved method in this work that models the knowledge base, named as Hierarchical Enhanced Colored Petri Net (HECPN). HECPN is a high-level Petri Net that has some additional features on Colored Petri Net and with which the modularised knowledge base can be modeled hierarchically. In this work, we analyze the reachability in the knowledge base modeled with HECPN and propose the methods to detect anomalies. Using these methods, both anomalies in the submodules and inter-module anomalies can be detected. Finally, we apply the proposed methods to the knowledge base verification for a nuclear power plant expert system to demonstrate their utility in this work. (C) 1999 Elsevier Science Ltd. All rights reserved.