Why is a Joint Prognostic Model with a Change Point (JPM-C) critical for analyzing internal resistance growth during battery degradation testing? Unlock Accurate Battery RUL Predictions

A Joint Prognostic Model with a Change Point (JPM-C) is critical because battery resistance growth is often nonstationary: it may remain gradual for much of a cell’s life and then accelerate sharply as degradation progresses. A conventional JPM assumes one continuous degradation relationship, so it can misestimate cell-specific behavior and produce unreliable remaining useful life (RUL) predictions after acceleration begins. JPM-C addresses this by estimating an unknown change point—using the Concordance Correlation Coefficient (CCC)—and modeling degradation before and after that transition.

The central value of JPM-C is that it aligns the prognostic model with the battery’s actual degradation phases. By combining continuous internal-resistance measurements with failure-time information and an estimated acceleration point, it can update individual-cell failure risk and RUL more realistically.

Why Internal Resistance Requires Change-Point Modeling

Resistance growth is not uniform across battery life

Internal resistance is often relatively stable during the early and middle portions of operation. As cycling continues, however, degradation mechanisms such as separator deterioration, electrode structural damage, and electrolyte loss can cause resistance to increase more rapidly.

During discharge, resistance can also rise sharply near deep discharge as active electrode materials become less conductive. These transitions demonstrate why a single fixed-rate degradation curve may not describe the full operating history.

The acceleration phase contains important prognostic information

A rapid increase in resistance is not merely another measurement trend. It can indicate that the cell has entered a more advanced degradation regime in which failure may approach more quickly.

If a model averages the early slow-growth period with the later rapid-growth period, it may dilute the significance of the acceleration. The resulting forecast can be especially misleading when the cell is already near a practical or defined failure threshold.

What JPM Contributes to Battery Prognostics

It combines condition monitoring with failure data

A Joint Prognostic Model links a continuous condition-monitoring signal—such as internal resistance—with time-to-event failure information. This allows the model to use both how the resistance is changing and when cells fail.

The resistance trajectory is typically modeled with a linear mixed-effects structure. Fixed effects describe population-level behavior, while random effects capture differences between individual cells.

It represents cell-to-cell variability

Cells produced with differences in electrode pressing, assembly, or material distribution may degrade at different rates. JPM accounts for this through cell-specific random parameters rather than assuming every cell follows the same trajectory.

A representative resistance model may be expressed as:

[ CM_i(t)=b_{i0}+b_{i1}t+b_{i2}t^2+e_i ]

Here, (b_i) represents the individual cell’s random effects, and (e_i) represents measurement noise.

It updates predictions as new measurements arrive

Within a Bayesian framework, JPM updates the individual cell’s parameter estimates as its resistance history grows. The updated condition information is then used as a time-dependent covariate in a Cox proportional hazard model.

This enables researchers to estimate an individual cell’s survival probability, failure risk, and remaining life without re-estimating the entire population model at every observation.

What the Change Point Adds

It separates distinct degradation regimes

JPM-C introduces an unknown change point that divides the resistance trajectory into at least two meaningful phases:

1. Pre-change-point behavior: relatively gradual or stable resistance growth.
2. Post-change-point behavior: accelerated resistance growth and potentially faster progression toward failure.

This structure is more consistent with the physical pattern often observed in extended degradation testing.

It prevents early-life data from dominating late-life predictions

Early measurements may contain limited prognostic information because resistance varies little across cells. If these observations are treated as representative of the entire life cycle, they can dominate the fitted trend and obscure late-stage acceleration.

By identifying the transition, JPM-C allows the model to give appropriate weight to the regime in which resistance begins changing more rapidly.

It uses CCC to identify the most informative transition

The Concordance Correlation Coefficient evaluates agreement between observed and modeled behavior. In the JPM-C framework, CCC can be used to estimate the change point that best aligns the model with the measured resistance trajectory.

The practical objective is not simply to locate a visually dramatic bend. It is to identify a transition that improves agreement between the model and the observed degradation behavior.

Why JPM-C Improves RUL Analysis

It reduces systematic prediction error

A fixed degradation curve can understate risk after resistance acceleration begins. Conversely, fitting a steep late-life trend to the entire history can overstate early-life degradation.

A change-point model reduces both types of distortion by assigning different behavior to the relevant stages of cell life.

It improves individual-cell risk estimation

Battery populations contain substantial cell-to-cell variation. JPM-C combines the population-level change-point structure with cell-specific random effects, allowing predictions to reflect both the general degradation pattern and the observed history of a particular cell.

This is important in research settings where the objective is not only to describe an average cell, but also to identify which individual cells are approaching failure.

It supports earlier recognition of accelerated degradation

Once the model detects that a cell has entered the post-change-point regime, its future resistance behavior can be assessed using the accelerated phase rather than the earlier stable phase.

That provides a more defensible basis for maintenance decisions, test termination, formulation comparison, and cycle-life estimation.

Understanding the Trade-offs

Change-point estimation can be uncertain

The transition may not occur at exactly one sharply observable cycle. Measurement noise, operating conditions, and cell variability can make the estimated change point uncertain.

Therefore, the change point should be treated as a model estimate supported by the data, not as an unquestionable physical boundary.

A more flexible model requires more careful validation

JPM-C has more structure than a standard JPM. Researchers must verify that the added change point improves predictive performance rather than simply fitting random fluctuations.

Validation should examine out-of-sample RUL accuracy, failure-probability calibration, and the stability of the estimated transition across cells and test conditions.

Internal resistance is not the right primary indicator for every application

For energy-storage and backup applications, capacity fade may be the more relevant health indicator because usable stored energy determines service performance. Internal resistance is particularly important for applications such as automotive starting batteries, where instantaneous power delivery is critical.

JPM-C should therefore be applied to the health indicator that best represents the intended failure definition.

Physical interpretation still requires engineering judgment

A statistical change point identifies a change in the observed degradation pattern. It does not, by itself, prove which physical mechanism caused the transition.

Electrochemical analysis, teardown studies, or complementary measurements may be needed to connect the statistical transition with specific material or structural degradation processes.

Making the Right Choice for Your Goal

JPM-C is most valuable when the test data show a credible transition from gradual to accelerated resistance growth.

• If your primary focus is accurate RUL prediction: Use JPM-C to distinguish early-life resistance behavior from the accelerated late-life regime before forecasting remaining life.
• If your primary focus is individual-cell screening: Combine the change-point structure with cell-specific random effects to identify cells whose resistance trajectories indicate elevated failure risk.
• If your primary focus is comparing battery designs or formulations: Apply the same change-point and failure definitions consistently across test groups, then compare transition behavior and post-change-point degradation.
• If your primary focus is energy-storage reliability: Consider modeling capacity fade as the primary health indicator, using resistance as a complementary diagnostic signal.
• If your primary focus is automotive or high-power performance: Prioritize resistance or impedance growth because increasing resistance directly affects power delivery.

JPM-C gives battery researchers a more faithful link between measured degradation behavior and the decisions that depend on it.

Summary Table:

To ensure accurate battery degradation analysis and RUL predictions, leverage advanced testing equipment from KINTEK. Our comprehensive laboratory solutions support battery R&D and materials research, from cell fabrication to testing. Contact us today to discuss how our equipment can enhance your research.