Session: 13-01 - Heat Exchangers

Paper Number: 126786

126786 - Multidisciplinary Optimisation of Additive Manufactured Heat Exchangers for Aeronautical Applications 

Topology Optimization (TO) is the computational process of determining solid distribution inside a given design region in order to maximize the value of an objective function under given constraints. TO of heat exchangers allows a greater design freedom as opposed to standard shape optimization, resulting in designs with improved performances and lightweight. However, due to the complexity of the resulting geometries, standard manufacturing techniques cannot be readily employed. Instead, Additive Manufacturing (AM) provides a flexible pattern for the manufacturing of these complex shapes. A promising application which combines both TO and AM, is the design of Micro-Channel Heat Exchangers (MCHX) with a lattice-like geometry. In MCHX, the network of micro-channels is generated by periodic repetition of a single unit (lattice cell) along the three spatial directions. The topology of the network can be varied by modulating (locally) the shape and parameters of lattice units, resulting in a lattice matrix characterized by improved fluid perfusion and high surface-to-volume ratio. The goal of the present work is to provide a proof-of-concept workflow for the topology optimization of additive manufactured MCHX. The target application is a next-gen cooling system for electric flight, where MCHX  are the key components in order to control working temperatures of batteries and fuel cells and to prevent overheating of on-board electronics.

TO of MCHX is intrinsically a multi-scale problem as the performances of the heat exchanger (macroscale) is strongly dependent on the fluid regime and geometry of micro-channels (microscale). In order to render this problem tractable in a time frame compatible with industrial needs, the proposed approach is based on homogenization theory. The lattice matrix is modeled as an equivalent porous medium (simulated at the macroscale in Immerflow, a CFD software by Optimad), where permeability tensor and heat transfer coefficient are assumed to be non-linear functions of flow conditions and local lattice geometry. Closure relationships are provided by a Machine Learning (ML) model, trained to infer permeabilityand heat transfer coefficients realizing a two-ways coupling between macro and micro-scale. The database for the training is provided by Direct Numerical Simulations (DNS) conducted for varying operating conditions at the microscale for several lattice cell geometries using OpenFOAM (solver buoyantPimpleFoam).

The proposed test case is provided by Rolls-Royce and represented by a MCHX enclosed within a solid case. Six heating plates are used to model the heat flux coming from a power electronic equipment. A water-glycol solution is used as the coolant fluid to control the maximum temperature of the solid phase. The goal of the optimization is to maximize heat rejection, i.e. maximize the fluid temperature at the outlet. Constraints are set on the maximum solid temperature and pressure drop across the device. The optimization workflow was implemented in modeFRONTIER (ESTECO). The optimization algorithm selected for this application is a hybrid algorithm which combines a steady-state Genetic Algorithm (GA) with a Sequential Quadratic Programming (SQP) optimizer. The resultant optimized geometry is characterized by an uneven distribution of channel diameters resulting in a significant increase in the tortuosity of the flow path. Increase in tortuosity leads to an increase of the average residence time of the coolant, which has a beneficial impact on heat removal. The increased tortuosity also results in a larger pressure drop and large temperature gradients in both the solid and the fluid phase. The optimized configuration is characterized by an increase in the heat rejection of 27% with respect to a uniform lattice and an increase of the pressure drop by a factor 15, which, however, is still largely below the feasibility limit.

In the full paper, we will compare the performance of the topology optimized HEX with experimental data of the ALM printed geometry. Total pressure loss across the channel, average and peak temperature at the surface will be compared to the hi-fidelity CFD results of the same geometry.

Presenting Author: Alessandro Alaia OPTIMAD

Presenting Author Biography: Dr Alessandro Alaia is a senior Research Engineer at the Optimad Engineering SRL, Based in Turin, Italy. A;essandro has a PhD in Mathematics and Engineering Sciences, Politecnico di Torino (2008-2011), he also holds a degree in Aeronautics and astronautics from the same university (2005-2006). Alessandro has been in Optimas Engineeing Srl for almost 13 years.

Authors:

Alessandro Chiodi Politecnico di Torino
Alessandro Alaia OPTIMAD
Edoardo Lombardi OPTIMAD
Marco Cisternino OPTIMAD
Konstantinos Gkaragkounis OPTIMAD
Shahrokh Shahpar Rolls-Royce Plc