Predictive value of degradation modes for electrode-resolved DC resistance in lithium-ion batteries

In lithium-ion batteries, capacity fade and resistance growth are coupled at the full cell (FC) level, but whether capacity-related degradation information predicts resistance at individual electrodes remains unclear. Degradation mode analysis yields candidate predictors by quantifying loss of lithium inventory ( LLI ) and loss of active material at the negative ( LAM NE ) and positive ( LAM PE ) electrodes. Their predictive value for electrode-resolved DC resistance is evaluated using 18 three-electrode cells with nickel-rich cathodes and graphite anodes aged under six calendaric, cyclic, and fast-charging protocols. Residualized permutation tests show that degradation modes contain resistance information beyond checkup number and capacity-based state of health for both electrodes. Support vector regression is then used for feature set ablation, aging path generalization, and temperature extrapolation. Feature set ablation shows that degradation modes improve prediction beyond operating-condition baselines, with greater gains for the positive electrode (PE) than the negative electrode (NE). Aging path generalization suggests that transfer is constrained by the joint degradation mode and resistance space represented in the training data. Similar LAM PE can accompany different PE resistance growth across aging paths, while LLI helps distinguish these states. FC resistance features restore PE transfer accuracy and improve NE prediction. For temperature extrapolation, a learnable Arrhenius layer reduces the RMSE at unseen temperatures by up to 66 %. Degradation modes therefore carry predictive information about electrode-resolved resistance, but the mapping is electrode-asymmetric and path dependent. Combining them with FC resistance and physically informed temperature modeling provides a more reliable basis for non-invasive electrode-level power fade diagnostics.