Community Detection in Flow-Based Engineering Networks

Different from the conventional structural perspective to detect communities, flow-based communities are more relevant to community functions and behaviors. This paper defines and detects flow-based communities based on Coupling Strength (CS) and corresponding functional modularity to understand community structures in engineering networks. In general, CS should be defined to reflect flow transmission ability between couples of nodes by considering specific physical and control characteristics in specific engineering fields. Then, the conventional modularity could be upgraded as functional modularity based on CS. The functional modularity could be applied as a metric to evaluate any network partitioning in a corresponding engineering field. Furthermore, the Newman fast algorithm is modified with maximization of functional modularity as objective. Cohesion tightness for a specific community is proposed so that communities with different scales in different networks can be quantitatively compared. Based on studies on some real networks from different categories, it is found that flow-based community and conventional topological communities may have considerable differences in variation rule, the magnitude of modularity, and community uniformity.