Battery Intelligence: Understanding the Signals Before Battery Failure

Batteries rarely fail without warning. They just warn us quietly. Before a cell overheats, it may start behaving differently. It may lose charge faster at rest, run slightly warmer than neighbouring cells, or show a small change in its voltage behaviour. A conventional battery management system (BMS) or battery monitoring system can detect many abnormal conditions, but subtle changes can still go unnoticed until they become a serious problem.

At SimLionics, we focus on those quiet signals. By combining physics-informed AI, battery modeling, battery simulations and battery digital twin software, SimLionics helps identify abnormal battery behaviour earlier, turning complex battery data into actionable intelligence.

Case Study 1: Detecting Cell Defects Through Battery Testing

Industry: Consumer Electronics
Scale: More than 1.6 million units recalled in the US

In June 2025, a leading consumer electronics brand recalled more than 1.1 million power banks following reports of fires and explosions. A second recall involving approximately 481,000 units followed in September. The company traced the problem to lithium-ion cells from a single supplier, identifying the issue through expanded component-level audits and supplier testing.

The Warning Signs

A cell defect can manifest through subtle changes in behaviour, including elevated self-discharge, where a cell loses charge while not in use, or unexplained heating during charging.

These behaviours can become important inputs during battery testing and early screening. Rather than looking only for an obvious failure condition, a physics-based baseline can help identify cells whose behaviour is beginning to deviate from expectations. Screening battery cells against physics-based behavioural baselines can help identify potential supplier or manufacturing issues before affected products reach customers.

Case Study 2: Battery Modeling Beyond Conventional BMS Detection

Industry: Electric Vehicles
Scale: Approximately 320,000 vehicles in the US in the latest recall

In November 2025, a major automaker recalled approximately 320,000 plug-in hybrid SUVs over battery fire risks and instructed owners not to charge the vehicles and to park them outdoors. The likely cause was separator damage inside the cells. The separator is the thin layer that keeps the positive and negative sides of a cell apart. Damage to it can create an internal short circuit. The important detail is that an earlier recall had already introduced BMS software designed to detect battery problems. However, the automaker found that the software was ineffective at detecting certain abnormalities, with fires reported in vehicles that had already received the software update. A different automaker experienced a similar situation with plug-in hybrids affected by cell separator defects, leading to another recall after cells vented in vehicles outside the earlier recall population.

The Warning Signs

A damaged separator can create a soft short, a small internal electrical leak that gradually drains the cell. This may appear as cell-to-cell voltage divergence or abnormal voltage relaxation after charging. This is where conventional battery management systems can benefit from deeper battery modeling. Instead of asking only whether a measurement has crossed a predefined threshold, a model can help determine why a particular cell is behaving differently from its neighbours. This highlights the need for battery detection to go beyond fixed thresholds. Understanding the underlying behaviour of a battery can provide a deeper layer of insight into developing abnormalities.

Case Study 3: Why Energy Storage Systems Need Cell-Level Battery Intelligence

Industry: Energy Storage Scale: A 300 MW facility in California

In January 2025, a fire broke out at one of the world's largest battery storage facilities in Moss Landing, California. The fire burned for days and flared up again a month later. More than 18 months later, the root cause remains unknown. In September 2026, approximately 1,200 battery modules that were still charged and could not be reached caught fire. Nearby residents were instructed to shelter in place. For operators of an energy storage system, this highlights a critical challenge: without detailed cell-level health data, understanding what happened and determining which remaining modules may still be at risk become significantly more difficult.

The Warning Signs

We cannot say exactly what the warning signs were in this case. And that is precisely the point. Without detailed cell-level health data, investigators have less information with which to reconstruct the sequence of events. Operators also have less visibility into the condition of individual modules. For a battery energy storage system (BESS), continuous cell-level monitoring can therefore serve a role beyond simply identifying a developing failure. Detailed battery data can help operators not only identify abnormal behaviour earlier, but also understand failures after they occur and make better-informed decisions about recovery and maintenance.

What the Three Cases Have in Common

The three cases, power bank recalls, plug-in hybrid recalls and grid-scale battery storage fires, all point to the same underlying challenge. In the power bank case, the issue began with a cell defect, while the plug-in hybrid recalls involved cell separator damage. In the grid storage case, the root cause remains unknown. Basic checks did not fully identify these issues, highlighting the need for approaches such as batch screening, physics-based detection and continuous cell-level monitoring. Across consumer electronics, EVs and energy storage, the pattern is similar: small changes at the cell level can precede much larger failures. This makes battery intelligence less about simply collecting more data and more about understanding what that data is actually telling us.

How SimLionics Helps

Physics-Informed AI

Physics-informed AI combines machine learning with an understanding of battery behaviour. Rather than treating battery AI as a black-box prediction layer, physics-informed approaches can use knowledge of battery behaviour to help identify abnormal patterns, reduce unnecessary alarms and provide greater insight into potential failure mechanisms. The objective is not simply to predict that something may go wrong, but to understand what is changing and why.

Battery Modeling & Simulations

Battery modeling helps represent how cells and batteries behave under different operating conditions. Combined with simulations, it can allow teams to evaluate battery behaviour, charging strategies and potential failure scenarios before they occur in the physical system. This can also support battery optimization by helping teams understand how operating conditions, thermal behaviour and charging strategies may influence battery performance and degradation.

Battery Digital Twin Software

Battery digital twin software creates a dynamic virtual representation of a physical battery. So, what is a digital twin? In battery applications, a digital twin is a virtual representation that uses operational data to reflect the condition and behaviour of its physical counterpart. This is where digital twin technology becomes particularly valuable for battery applications. By continuously incorporating battery data into the virtual model, a digital twin can help track parameters such as State of Health (SoH), forecast Remaining Useful Life (RUL) and identify cells that are drifting away from expected behaviour. Together, these capabilities create a more intelligent approach to monitoring and managing batteries throughout their operational life.

From Monitoring to Predictive Battery Maintenance

Traditional battery monitoring answers an important question: “What is the battery doing right now?” A more intelligent approach can go further: “How is the battery changing, why is it changing, and what could happen next?”

This shift is at the heart of predictive battery maintenance. Instead of waiting for a battery to cross a critical threshold or experience a failure, organisations can use battery intelligence to identify changing behaviour earlier and determine when intervention may be required.

This can support:

• Earlier identification of abnormal cells
• More informed maintenance decisions
• Better battery testing and quality assessment
• Reduced uncertainty around battery condition
• Improved operational planning
• More informed replacement and retirement decisions

Supporting Battery Lifecycle Management

Battery intelligence also has value beyond immediate failure prediction. For manufacturers, insights from battery data can support testing, quality decisions and product development. For fleet operators, they can help identify batteries that are degrading differently from expected. For energy storage operators, they can support maintenance planning and safer asset operation. This creates a broader role for battery lifecycle management, understanding battery behaviour from testing and deployment through operation, maintenance and eventual replacement. The goal isn't simply to generate more dashboards.

It is to answer practical questions:

• Is this battery ageing normally?
• Which cells are behaving differently?
• What is likely to happen next?
• When should the battery be inspected?
• Should its operating or charging strategy change?
• When should it be repaired or retired?

Conclusion

Power banks, plug-in hybrids and grid-scale storage may serve very different applications, but the underlying challenge is similar: Small cell-level problems can develop before conventional monitoring identifies a serious failure. The opportunity is to detect and understand those changes earlier. By combining battery intelligence, physics-informed AI, battery modeling, simulations and battery digital twin software, SimLionics helps turn complex battery behaviour into actionable insights. The result is a shift from reactive monitoring towards predictive battery maintenance, supporting safer operation, better decisions and more effective battery lifecycle management.

Want to bring early-warning intelligence to your batteries?

Connect with the SimLionics team to explore physics-informed battery modeling, predictive battery maintenance and battery digital twin solutions for your application.