The use of computational fluid dynamics (CFD) to ascertain the performance of thermoelectric generators (TEGs) coupled to waste-heat sources has become more prevalent, whether in a segregated or fully-coupled fashion. The inherent dependence of TEG performance, both thermally and electrically, is predicated upon the ability to establish a temperature difference across the device. The importance of modeling the waste heat stream, as well as heat exchange geometry, is paramount in faithfully predicting device performance. The use of Reynolds-averaged Navier-Stokes (RANS)-based equations in CFD models is invariably implemented due to the relatively low computational cost. Although RANS-based models allow for the quick evaluation of a myriad of designs, their use is not without limitation. To provide closure to the Reynolds stress tensor and turbulent heat flux, each RANS model introduces assumptions and simplifications in predicting the effects of turbulence on momentum and heat transfer. Without a set of experimental validation data to benchmark existing segregated or coupled thermal-fluid-electric models against, accompanied by the abundance of applicable RANS-based models, there exists great uncertainty in the use of CFD as a predictive tool for the evaluation of permissible device designs. To this end, a fully-coupled thermal-fluid-electric numerical model of a novel TEG is introduced, and the process of quantifying model form uncertainty, as well as discretization uncertainty, is demonstrated. Various RANS-based models are used in the evaluation of device performance, and insight into the relative required fidelity of the RANS-based models to dependably predict performance is provided.
Video presentation