Abstract The ATLAS Technical Coordination Expert System is a knowledge-based application which describes and simulates the ATLAS experiment based on its components and their relationships with differing levels of granularity but with an emphasis on general infrastructure. It facilitates the sharing of knowledge and improves the communication among experts with different backgrounds and domains of expertise. The developed software has become essential for the planning of interventions as it gives easily insight into their consequences. Furthermore, it has also proven to be useful for exploring the most effective ways to improve the ATLAS operation and reliability by identifying points of failure with significant impact. The underlying database describes more than 13,000 elements with 89,000 relationships among them. It combines information from diverse domains such as detector control and safety systems, gas and water supplies, cooling, ventilation, cryogenics, and electricity distribution. As the most recent addition, a tool to identify the most probable cause of a failure state has been developed. This paper discusses the graph-based algorithm currently implemented by that tool and shows its behaviour based on the parameters entered by the user. An example in form of a real failure event is given which demonstrates the potential of the Expert System for understanding major failures faster in urgent situations.