B1295
Title: A longitudinal set-up for degradation modelling
Authors: Maria Kateri - RWTH Aachen University (Germany) [presenting]
Abstract: The study of aging is of special interest for many products and devices, e.g., lithium-ion batteries (LIB). A stochastic modelling framework is proposed for product aging, dealing with a special experimental framework under which each item of a sample under test is measured repeatedly over time, providing a sequence of values but with a small number of observations that do not allow for standard degradation modelling. In a longitudinal setting, a class of linear mixed effects models are introduced for describing the degradation paths in case of an aging experiment with sparse data. The aim is to assess the aging paths, enabling joint consideration of multiple experimental conditions through a condition-based grouping of the data and allowing for individual (random) effects corresponding to different initial levels. After introducing the model and discussing parameters estimation and goodness of fit testing, the model is applied to experimental data of LIB cells. Next, a new procedure for simulating experimental data is proposed based on this model that can be used for data augmentation or simulation-based inferential procedures, significant in case of data sparsity. Finally, the robustness of such models against misspecification of tuning parameters is assessed by a simulation study.
