PhD Defence Citra Vidya | Modeling and adjoint optimization of heat exchanger geometrie

MODELING AND ADJOINT OPTIMIZATION OF HEAT EXCHANGER GEOMETRIE

Citra Vidya is a PhD student in the research group Thermal Engineering. Her supervisor is prof.dr.ir. T.H. van der Meer from the Faculty of Engineering Technology.

In today's society, with an increasing need for energy worldwide and at the same time a growing awareness of its negative consequences, optimizing and reducing energy consumption has become important for sustainability. The process of optimizing the appliances in large industry and household scale contributes to the minimization of the global energy consumption. In order to increase the efficiency of these appliances, various optimizations are employed, such as the shape optimization.

In this thesis, a shape optimization method is used to modify the geometry of an existing heat exchanger. The heat exchanger is used in a domestic boiler and the shape optimization is a crucial step for its fast development. Currently, the shape of the heat exchanger is highly dependent on its manufacturing process. Thus, more complex shapes are investigated and their effects on the heat transfer and pressure drop are evaluated in this thesis. Two aspects need to be investigated for this heat exchanger optimization: the complex internal flow within the heat exchanger, and the performance of the optimization method as well as its resulting shapes. Since the heat exchanger consists of different flow regimes due to its complex geometry, a detailed study on these regimes is conducted prior to the optimization procedure. Furthermore, the complex geometry of this heat exchanger leads to different models used in this thesis. Therefore this thesis is divided into three parts based on the models: the three-dimensional single cylinder, the two-dimensional cylinder arrays, and the three-dimensional cylinder array.

In optimizing the heat exchanger, the shape optimization method is justified as the most suitable method. Using the ANSYS Fluent's adjoint shape optimization method, the pressure drop and the heat transfer of the heat exchanger are optimized. This method has been used in the past for optimizing various geometries, yet to the author's knowledge, no detailed study on the cylinder array has been published. Consequently, a study on the adjoint parameters and its effect on the overall performance of the optimization procedure is of importance.

To investigate these aspects, three sets of studies are conducted, and are presented in the three parts of the thesis. Firstly, an unsteady Direct Numerical Simulation (DNS) of a three-dimensional single cylinder in a cross-flow is performed at critical Reynolds number, ReD = 2000. This flow is categorized as a transition flow in shear layer regime, a regime that is not widely studied in literature aside from detailed numerical simulations at ReD = 3300 and ReD = 3900. The findings of this work provide an insight into the heat transfer over a circular cylinder in a transitional flow. Due to the periodic flow and the three-dimensional motions at this Reynolds regime, further studies using the single cylinder domain are conducted at lower Reynolds number, ReD = 10. At this flow regime, a steady adjoint shape optimization procedure is performed with a conjugate heat transfer model. The optimization results produce the optimized shape of a single cylinder as well as a reduction of the drag force and an improvement of the heat transfer.

The second part of the thesis consists of two-dimensional studies of cylinders in an array. A few studies on the shape optimization of cylinder array have been presented in literature, however none of these studies employ the adjoint method. In this part of the thesis, four topics are elaborated: the first deals with detailed single-objective and multi-objective optimization cases where pressure drop and heat transfer are chosen to be the objectives. The results of these cases provide insight into the correlation between the heat transfer and the pressure drop of an array of cylinders for this specific flow. Second, both circular and elliptical cylinders are modeled and optimized. The final optimized shapes of both cases are compared and their performances are visualized in the so-called objective space in terms of heat transfer and pressure drop. This study yields an overview of the objective space and the possible optimization paths for these cases. Third, various adjoint parameters are studied and its effects on the final optimized geometries are shown. Finally, the challenges that arise when applying the ANSYS adjoint automatic shape optimization procedure are explained and the solutions are presented.

In the last part of the thesis, a three-dimensional array of cylinders is modeled. A more complex boundary condition is used, namely the periodic boundaries at the inlet and at the outlet of the domain. The conjugate heat transfer model is employed and the domain is optimized for a flow at ReD = 100. The optimization results show that the adjoint optimization procedure cannot successfully produce a shape that yield improvements in both objectives, alternately optimizing only for one objective. Following this, the investigation of the failure of the adjoint optimization procedure is presented. Studies on the weight factors and optimizer performance are conducted and the results are compared with that of the successful optimized single cylinder case. Possible cause of the failure of the adjoint optimization procedure is presented as well as the best practice and recommendations for future work.