Considerations on the thermal performances of ribbed channels by means of a novel dynamic method for hierarchical clustering

The investigation of ribbed surfaces for the enhancement of heat transfer in forced convection allowed to observe that different geometries may lead to comparable performances. Due to the lack of an underlying structure of the data, a novel method for data clustering is introduced here, to assess to what extent comparable performances can be achieved using different rib geometries. The clustering method is an agglomerative technique, based on the inclusion of each configuration in another ones bounding box, whose size depends dynamically on the Nusselt number and the pumping power. The method is applied to a large database experimentally obtained at ThermALab of Politecnico di Milano, in order to identify the Nusselt number and the friction factor for diverse-rib configurations in a large-aspect ratio channel with low-Reynolds flows. The clusters are determined, and the resulting families of configurations are used to assess the possible effects of the rib eometry on the thermal and fluid-dynamic performances. The clustering analysis results suggest interesting considerations.