Design of a class E inverter with stabilized output power using artificial neural network for applications in biomedical implants

This paper presents the design, simulation, and experimental validation of a load-independent class E inverter tailored for biomedical implant applications. The proposed system addresses the challenge in the PID controller of maintaining constant output power without relying on conventional feedback circuits, which often face difficulties in accurately sensing load resistance, especially in implantable environments. To overcome this, a dual artificial neural network (ANN) architecture is introduced. The primary ANN serves as the main controller, while the secondary ANN estimates the load resistance using only the DC input voltage of the inverter and average input current. The estimated resistance is fed into the primary ANN, which regulates the system to ensure stable power delivery regardless of load variations. The inverter receives its required DC voltage from a buck converter, whose duty ratio is adjusted by the primary ANN with a mean relative error below 0.01 to maintain constant output power. Theoretical analysis, simulation, and experimental results confirm consistent performance, achieving 2 W output power at 1 MHz switching frequency. The compact and feedback-free nature of this design makes it well-suited for wireless power transfer in medical implants, wearable electronics, and portable consumer devices.