SYDE Graduate Seminar - Closed-Loop Electrochemical Impedance Spectroscopy Using Fisher Information Optimization & Geometric Identifiability Analysis

Presenter

Yasaman Masoudi, PhD candidate in Systems Design Engineering

Abstract

Electrochemical Impedance Spectroscopy (EIS) is a powerful tool widely utilized in engineering and scientific disciplines for system characterization and testing. In classic or conventional EIS, excitation is performed as an open-loop system over uniformly distributed frequencies in the logarithmic scale without incorporating any feedback from the preceding steps. The number of frequencies is chosen based on trial and error, which in turn could add to the time and cost of the EIS tests. To address the gaps in classic EIS, this work proposes a practically implementable closed-loop EIS framework based on Fisher information matrix (FIM) determinant-optimization problem, where the excitation frequencies are iteratively chosen to maximize the sensitivity of electrochemical based equivalent circuit models (ECMs) to estimation parameters given the prior observations and estimation data. The proposed FIM EIS framework includes termination criteria based on successive convergence of the estimation parameters. Various FIM EIS formulations are elaborated, and benchmarked against classic EIS using uniform sampling. Additionally, another benchmarking system based on sequential EIS with mixed uniform-random sampling technique is proposed, and the results are compared with those of the FIM and classic EIS methods. The effectiveness of FIM-based EIS is extensively evaluated and compared through synthetic and experimental data and various ordinary- and fractional-order ECM models. The results confirm that FIM EIS methods collect samples from locations of maximum information, which can reduce the time and cost of EIS tests without compromising estimation accuracy. Moreover, by using the geometric identifiability analysis, it is shown that the classic FIM EIS leads to redundant samples. A geometric FIM EIS is then proposed, which reduces the number of sampling regions without compromising the estimation accuracy.

Join virtually via MS Teams!

This seminar counts towards the graduate student seminar attendance milestone!