In previous work the authors proposed a distributed diagnosis approach consisting of two phases—preliminary diagnosis in each local diagnoser and inter-diagnoser communication. The objective of communication is to achieve either global or local consistency among local diagnoses, where global consistency is captured by the equilibrium concept of supremal global support. To achieve this equilibrium, an algorithm called Computational Procedure for Global Consistency (CPGC) was proposed. But it turns out that CPGC has high time complexity and weak scalability. To rectify these shortcomings, we propose a hierarchical computational procedure. A further advantage of this procedure is demonstrated, based on multiresolutional diagnosis. With the latter, fault detection is conducted at each hierarchical level, so that computation can be confined to those modules likely to possess faults, while fault-free modules are safely disregarded. A simplified industrial example is provided in illustration.