Why temperature-only warning struggles in liquid-cooled packs
Lithium-ion batteries power EVs, stationary storage, and an expanding range of industrial equipment, but thermal runaway remains the central safety concern. Once a reaction becomes self-sustaining, a single-cell failure can escalate into a system-level event within seconds.
Temperature has long been the most common early-warning signal: sensors are mature, low-cost, and easy to integrate. The limitation is that temperature is a macroscopic response to heat accumulation. In the early stage of instability, it often lags the underlying physical and chemical processes — exothermic side reactions, gas generation, structural swelling, and electrochemical anomalies.
This gap widens with immersion cooling. By submerging cells in electrically insulating fluid, immersion designs remove heat faster and more uniformly, and can suppress local overheating. But the same cooling that holds peak temperature down also flattens the surface temperature rise rate. The temperature signal becomes smoother and slower to move, so a fixed temperature threshold triggers later and narrows the response window. The better the cooling, the later temperature-based warning can appear.
Looking at the interface instead of the bulk temperature
As cooling improves, engineers are looking for signals that are less affected by external heat exchange and reflect the internal progression of the cell. During thermal runaway, electrolyte decomposition and parasitic electrode reactions continue to generate gas, driving internal pressure buildup and housing swelling. These mechanical changes originate inside the cell rather than from external heat transfer, so they evolve with less dependence on the cooling environment and typically appear before a macroscopic failure event.
Recent published research monitored this mechanical response in situ using a thin-film pressure sensor integrated into the battery system. The structural response was not only an overall rise in load amplitude: the contact area expanded progressively, and the pressure distribution pattern kept reorganizing. Spatially resolved mechanical response captured precursors that overall structural indicators missed.
The work framed the early-warning descriptor at the interface level: the contact area A between sensor and cell housing, and its time derivative dA/dt, used as a low-temperature early-warning indicator. In the reported controlled thermal-runaway platform, the interface-pressure signal showed structural instability at a side-wall temperature of only 25.6 °C — roughly 29 minutes before the safety valve opened and 55 minutes before thermal runaway actually occurred.
What the thin-film sensor measures
The tested sensor was a CNSCAN-spec thin-film pressure sensor used for in-cell deformation monitoring:
- Thickness: 0.46 mm, thin enough to conform to the cell housing without significantly altering the original mechanical constraint
- Measurement points: 2,288 (density 6.9 points/cm²)
- Pressure range: 0–3,447 kPa
- Accuracy: 5%
- Pressure resolution: 1.2 kPa
The flexible multilayer structure uses double-layer PET as substrate and protective layer, followed by a silver paste conductive layer, a carbon paste pressure-sensitive layer, an adhesive layer, and an internal air-channel layer to enhance deformation sensitivity. Sensing elements are grid-pattern semiconductor units arranged in a crossed row-and-column matrix. Under load, the carbon microstructure deforms and changes resistance; after calibration, each node’s local pressure is quantified. A fast scanning circuit reads the resistance at every cross point to deliver real-time pressure amplitude and distribution.
The reported test cell was a 280 Ah commercial prismatic LFP cell (3.20 V nominal, 173.7 × 217.2 × 71.7 mm, 5,420 ± 30 g), charged to 100% SOC and heated on one side to trigger runaway, with aerogel insulation on both sides.
What the comparison showed
Under Accelerating Rate Calorimetry near-adiabatic conditions, the cell showed three thermal stages: thermal stability below 99.3 °C, accelerated self-heating from 229.1 °C, and a final peak of 425.6 °C.
Across three conditions — bare cell, Oil-A immersion, and Oil-B immersion — immersion delayed the runaway trigger and lowered the peak temperature, but it also flattened the temperature rise curve. The pressure-based interface descriptor tracked the internal mechanical progression independently of how strongly the coolant masked the surface temperature.
What this means for battery R&D and QC
- Add a mechanical channel beside temperature. Pressure distribution does not replace thermal, gas, voltage, or acoustic monitoring; it adds an interface-level signal that cooling cannot smooth out as easily.
- Watch distribution, not just peak force. Progressive contact-area expansion and pattern reorganization can precede a large amplitude change.
- Keep the sensor influence minimal. Sub-0.5 mm flexible construction is important so the monitoring hardware does not change the cell’s real boundary conditions.
- Use spatial data in validation. Map where pressure concentrates on the housing to support pack structural design and compression strategy review.
Explore CNSCAN pressure mapping for battery applications
CNSCAN supplies CN-series thin-film pressure sensors and distributed pressure measurement systems for new-energy battery R&D, formation and stacking pressure monitoring, and pack assembly quality control. Browse the pressure mapping sensors and pressure mapping systems, or see more industry applications. For a specific cell or pack setup, contact our team at contact@cnscan.com to discuss sensor sizing and data acquisition.
