On the Thermal Efficiencies of Cascading Heat Exchangers: An Algorithmic Approach - II

Abstract

A scalable design architecture for addressing optimal temperature performance on demand is investigated and implemented in this research using novel algorithmic methods. The thermal efficiencies of cascading multiple heat exchangers (HEX) in a parallel fashion networked to multiple cold plates is presented. An in-depth, self-learning algorithmic approach is presented to characterize power and performance perspectives. Several novel algorithmic approaches aimed at reducing transient thermal solution response time for coolant flow branching to dynamically and to preemptively address the Tjrise has been investigated and implemented in this design. The dynamic algorithm also calculates the efficiency of the system based on the heat picked up by the cold plate and heat dumped by the HEX upon recirculation and forced convectively cooled by the fans blowing on the HEX relative to the ambient temperature. The apparatus once deployed, is driven by algorithmic methods without the need for human intervention and the flow control is automated. Experimental data for quad HEX configuration dissipating 4 kW on multiple cold plates is presented. The algorithm automatically branches coolant to the cold plates only when power is being dissipated by the package under the cold plate. A plug and play, adaptive, coolant flow branching system that removes human intervention and dynamically optimizes thermal efficiency based on real-time, dynamic power changing conditions is implemented and presented in this research. When implemented in data centers, this is projected to optimize TCO (Total Cost of Ownership) at the rack level and for large scale deployment & adoption at data centers.