Optimizing model size in diagnosing complex aerospace systems

Describes a model-based approach for efficiently diagnosing complex aerospace systems with structural replication, i.e., the occurrence of multiple copies of components within a system. Model-based diagnostics provide the capability for high-fidelity, component-based diagnostics, but are typically both memory- and computation-intensive. We describe a component-based technique that reduces the memory and computation requirements of a model by taking advantage of structural replication without sacrificing diagnostic completeness or soundness. For example, in many systems it is common to have doubly- or triply-redundant sub-systems for improved system reliability. In other domains, a large system is composed of multiple replications of a small set of components. In either case, this component-based approach allows us to take advantage of the structural replication to produce a smaller model, which provide the same guarantees of diagnostic completeness and soundness as the traditional model.