Efficient design optimization of a miniaturized thermoelectric generator for electrically active implants based on parametric model order reduction.

This research focuses on the design of a miniaturized thermoelectric generator (TEG) for electrically active implants. Its design optimization is performed using the finite element method. A simplified TEG model is obtained by replacing the thermocouple array with a single representative thermopile, which considers the number and fill factor of the thermocouples as parameters. Instead of rebuilding the geometry of a detailed model with multiple thermocouples, the simplified model adapts the material properties of its representative thermopile, facilitating design optimization. We extend the model by integrating the simplified TEG together with a housing inside a human tissue model for thermoelectric analysis. For computation efficiency and applicability of model order reduction (MOR), a thermal model is derived from the thermoelectric one, with the Peltier effect being considered through an effective thermal conductivity. Through parametric MOR, two parametric reduced-order models are generated from the full-scale thermoelectric and thermal model, respectively. Furthermore, we demonstrate the design optimization of TEG both in full-scale and reduced-order model for maximal power output and sufficient voltage output.

System-Level Model and Simulation of a Frequency-Tunable Vibration Energy Harvester.

In this paper, we present a macroscale multiresonant vibration-based energy harvester. The device features frequency tunability through magnetostatic actuation on the resonator. The magnetic tuning scheme uses external magnets on linear stages. The system-level model demonstrates autonomous adaptation of resonance frequency to the dominant ambient frequencies. The harvester is designed such that its two fundamental modes appear in the range of (50,100) Hz which is a typical frequency range for vibrations found in industrial applications. The dual- frequency characteristics of the proposed design together with the frequency agility result in an increased operative harvesting frequency range. In order to allow a time-efficient simulation of the model, a reduced order model has been derived from a finite element model. A tuning control algorithm based on maximum-voltage tracking has been implemented in the model. The device was characterized experimentally to deliver a power output of 500 µW at an excitation level of 0.5 g at the respected frequencies of 63.3 and 76.4 Hz. In a design optimization effort, an improved geometry has been derived. It yields more close resonance frequencies and optimized performance.