On Applying Network Theory to Assembly Analysis

Abstract

The modelling of modern assemblies, particularly those in the aerospace industry, can be a challenging task. The complexity and size of these assemblies often make them difficult to manage and model. In addition to this, analysis techniques have little to no interoperability, driving the search for a one-size-fits-all solution. To address this problem, this paper presents a novel approach using graph networks to model assemblies. Three common assembly problems were used to test this proposal: identifying critical components, determining sub-assemblies, and analysing the critical tolerance paths. To address these assembly issues a tool was developed to convert an assembly model into a network, which is computationally lightweight and easy to analyse. Three families of network analysis techniques were then chosen to address the assembly issues: centrality measures, community detection algorithms, and a critical path algorithm. The results of this analysis showed that the network-based approach was able to automatically identify critical components through the centrality algorithms and propose sub-assemblies from the communities. The critical path algorithm was not able to clearly predict the most problematic variation stacks, yet with a tighter problem definition this approach is thought to be able to determine the largest variations between specific parts or within sub-assemblies. As such this study demonstrates the ability of network analysis methods to generate value across a variety of assembly challenges faced by industry.