An implementation of impedance spectroscopy approach to predict thermoelectronic properties of elements using a connectionist model of artificial neural network

Abstract Recent developments based on thermoelectric materials show their increasing potential in different applications. Starting from the recent experimental data available in the literature, we predict the thermoelectric properties of materials using an artificial neural network approach. The relevant thermoelectric parameters considered are the Seebeck coefficient, the thermal conductivity, the thermal diffusivity and the Peltier coefficient. The thermoelectric resistance and the characteristic frequency of thermal diffusion are also estimated by the analysis of the equivalent impedance of the interface studied. To build up our neural network model, we define the impedance approach of the thermoelectric system under adiabatic conditions by establishing the equation of heat over the chosen frequency domain. In addition, thermoelectric measurements by impedancemetry are applied to train the neurocomputing model. This approach can define the thermal properties of certain metals and semiconductors. We show the experimental results and those predicted by the corresponding model. Simulation results are also analyzed and compared to impedance spectroscopy measurements.