Sachem is an extensive large-scale real time knowledge-based system designed to monitor and diagnose blast furnaces. This paper aims at illustrating the way the concept of discrete event allowed the definition of a “perception-based diagnosis” approach as a recursive and holographic abstraction process of a discrete event. The example of the diagnosis of a “thermal load” phenomenon on a blast furnace is used in order to illustrate the way Sachem apply the “perception-based diagnosis” approach. Some considerations about the blast furnace and the development of Sachem are also presented in the paper to recall the complexity and the issue of the design of powerful perception systems.