Recursive Subspace Identification for Online Thermal Management of Implantable Devices

Abstract

This paper focuses on application of subspace identification methods to predict the thermal dynamics of bio-implants, e.g. UEA. Recursive subspace identification method implemented in this paper predicts the temperature readings of heat sensors in an online fashion within a finite time window and updates the system parameters iteratively to improve the performance of the algorithm. Algorithm validation is realized using COMSOL software simulations as well as using an in vitro experimental system. Both simulation and experimental results indicate that the proposed method can accurately predict the thermal dynamics of the system. The experimental results show online prediction of the thermal effect with a mean squared error of $1. 569 \times 10^{-2}$ °C for randomly generated Gaussian inputs and $3. 46 \times 10^{-3}$ °C for square wave inputs after adaptive filters converge.