This paper describes a deep-knowledge expert system shell for diagnosing faults in process operations. The expert program shell is called GOTRES (GOal TRee Expert System) and uses a goal tree- success tree deep-knowledge structure to model its knowledge- base. To demonstrate GOTRES, we have built an on-line fault diagnosis expert system for an experimental nuclear reactor facility using this shell. The expert system is capable of diagnosing fault conditions using system goal tree as well as utilizing accumulated operating knowledge to predict plant causal and temporal behaviors. The GOTRES shell has also been used for root-cause detection and analysis in a nuclear plant.