Neural Network-Based Design of Wireless Power Transfer Systems for Implantable Medical Devices

Wireless Power Transfer (WPT) is an essential technology for powering Implantable Medical Devices (IMDs). The significant distance between the transmitter and the receiver of the inductive link, relative to their small size, requires the use of High Frequencies (HF), hence increasing the complexity of the system design. This article proposes a method to model a HF inductive link based on Artificial Neural Networks (ANN), capable of estimating its electrical variables with an average error of 1.5% compared to Finite Element Analysis (FEA) data. This ANN is integrated into a multi-objective optimization process to enhance system efficiency under nominal conditions, taking into account electromagnetic safety assessment through the computation of the Specific Absorption Rate (SAR). This method is validated through the design of four WPT topologies, based on combinations of two inverters (Class D ZVS and Class E) and two rectifiers (Current-fed Class E and Compact Voltage-fed Class E). The designs are experimentally validated under various loads, distances, misalignments, and the use of biologic tissues as transmission media. All four topologies are capable of transmitting 0.48 W of power with efficiencies of 70% at a distance of 15 mm, achieving higher power density and efficiencies compared to state-of-the-art studies.