Multi-parameter design optimization of segmented annular thermoelectric generators: enhanced machine learning

The vehicle’s exhaust pipe, which is equipped with thermoelectric generators (TEGs), has been thoroughly studied for its potential to reclaim waste heat from the internal combustion (IC) engine. Nevertheless, the elevated temperature of exhaust gases could negatively impact the thermal reliability of the TEGs. This research focuses on assessing the thermal reliability and thermoelectric power output of unicouple and unileg segmented annular thermoelectric generators (SATEGs) that are mounted on the vehicle’s exhaust pipe. A numerical model was developed using ANSYS software, in conjunction with adaptive neuro-fuzzy inference system (ANFIS) prediction models, to optimize eight parameters. These parameters include the dimensions of the SATEGs, the heat transfer coefficient, and the temperatures of both exhaust gas and ambient air. In this model, skutterudite is utilized in the upper segment, whereas bismuth telluride is incorporated in the lower segment. The results reveal that the highest thermal stresses in the skutterudite unileg segment are below its yield stress when the air heat transfer coefficient exceeds 50 W/m²·K. In contrast, the bismuth telluride unileg segment exhibits thermal stresses below its yield stress at exhaust gas transfer coefficients of less than 85 W/m²·K. Furthermore, the unicouple and unileg SATEGs achieve maximum thermoelectric power outputs of 8.4 and 11.7 W at average temperature differences of 43.2 and 58.7 °C, respectively. Utilizing the ANFIS prediction model contributes to the optimization of various parameters, thereby enhancing the technical evaluation of extracting thermal energy from the IC engine and improving thermal reliability.