The exploitation of eigenspectra in electrochemical impedance spectroscopy: Reconstruction of spectra from sparse measurements

Electrochemical Impedance Spectroscopy (EIS) represents one of the most widely utilized techniques for the characterization of solid oxide cells (SOCs) and stacks in contemporary research. This work examines patterns in a set X of over 2,600 EIS measurements conducted on SOC stack level to identify a low- dimensional representation of the data. Besides the efficient training of data-driven models in such, this work primarily focuses on enabling the reconstruction of complete spectra from sampling the impedance at a limited number of relevant frequencies, thus significantly reducing the measurement time. By applying singular value decomposition, a reconstruction matrix can be developed containing a set of r patterns to sufficiently describe all the EIS in X . Based on these patterns, a set of r /2 tailored frequencies can be determined. For every permutation of measurement conditions in X and for r between four and twelve, the reconstruction results are comprehensively discussed utilizing EIS measurements conducted on a holdout SOC stack and thus not being part of the data set X. Furthermore, EIS containing varying faults and artificially distorted EIS are used to also capture unforeseen behavior. Depending on the measurement conditions accurate reconstructions can be reported for r = 6, i.e. three tailored frequencies.