A set of guidelines is described for generating an initial organization of knowledge for distributed diagnosis of a process plant. The diagnostic knowledge is organized hierarchically by primary processing systems (commonly feed, reaction, and separation in chemical plants), subsystems, components, behaviors and malfunction modes. The resulting classification hierarchy decomposes the diagnostic problem solving into coordinated, distributed modules, where different modules may use different methods to address specific local subproblems. Classification hierarchies, organized in this way, provide effective modularity for organizing large-scale, knowledge-based diagnostic systems, which are difficult to construct without pertinent organizing principles. Such hierarchies provide a framework for systematic knowledge acquisition and maintenance. Application of the guidelines emphasizes readily available sources of knowledge, considers common design and operating objectives of process plants, draws upon operating expertise and builds on generic process characteristics. Application is illustrated for a fluidized catalytic cracking unit and a paraxylene production unit.