Voltage Collapse and Early Failure Indicators in a Degraded EV Battery Under High-Current Load

Featured Application

The proposed methodology can be used in practice to identify degraded battery modules in aging EVs before critical failure occurs. By combining simple OBD-based diagnostics with targeted high-current testing, it becomes possible to detect cells that show abnormal voltage drops under load. This approach may be useful for vehicle service centers and battery recycling facilities, where quick screening of battery condition is required. It can support decisions on module replacement or second-life application, while reducing the risk of operating unstable battery segments.

Abstract

This paper investigates the safety behavior of degraded lithium-ion battery modules taken from a 2016 Nissan Leaf (30 kWh, 106,394 km). The vehicle exhibited typical failure symptoms, including P33E6 faults, sudden range drops, and activation of turtle mode under load. Initial diagnostics based on LeafSpy data revealed strong cell imbalance, with a voltage spread exceeding 2.3 V under high current (≈170 A). The weakest cells dropped close to 1 V, suggesting severe internal degradation. To better understand this behavior, selected modules (cells 73–88) were removed and tested under controlled laboratory conditions. Capacity measurements in a 16S2P configuration showed 49.8 Ah in the 4.1–3.1 V range, corresponding to a state of health of about 59%, which is consistent with BMS estimates. However, high-current discharge tests on the weakest segment revealed a much more critical picture. One cell experienced rapid voltage collapse (from ~4.0 V to ~1.2 V), accompanied by a sharp increase in voltage divergence and visible thermal effects. Infrared observations indicated localized heating up to 43 °C and irreversible swelling, pointing to early-stage electro-thermal instability. These results suggest that moderate SOH values do not necessarily reflect actual safety margins under dynamic load conditions. Overall, the study shows that simple OBD-based diagnostics can help identify problematic modules, but additional load testing is necessary to assess real safety risks in aged EV battery systems.

1. Introduction

The electrification of road transport is currently driven not only by technological progress but also by regulatory pressure and industrial policy [1,2]. In the European Union, the Fit for 55 package defines a clear pathway toward zero-emission passenger cars and vans by 2035. Similar policy directions are observed in other regions, including the United Kingdom and several non-European markets, where the expansion of electric vehicle (EV) fleets and charging infrastructure is accelerating [3,4]. As a result, increasing attention is being paid not only to vehicle deployment, but also to long-term battery durability, diagnostic capabilities during operation, and safety at later stages of the lifecycle, particularly as early-generation EVs begin to reach advanced aging stages in real-world use [5,6].

With many EVs now exceeding several years of operation, the focus is gradually shifting toward the reliability and safety of aged battery systems [7,8]. Early-generation vehicles, including widely deployed compact EV platforms from the mid-2010s, provide valuable insight in this context due to their long-term exposure to real driving conditions. After extended use—typically beyond 100,000 km and 8–10 years—these systems often exhibit noticeable degradation, manifested by capacity loss, increased internal resistance, and growing cell-to-cell voltage imbalance.

The mechanisms responsible for lithium-ion battery aging have been extensively studied, particularly in relation to solid electrolyte interphase (SEI) growth, lithium plating, and loss of active material [9,10]. These processes interact and lead to loss of lithium inventory, impedance rise, and gradual performance decline [11,12]. Degradation is therefore understood as a multi-physical phenomenon involving coupled electrochemical, thermal, and mechanical effects evolving over time [13,14].

Batteries based on nickel–manganese–cobalt (NMC) and lithium–manganese oxide (LMO) chemistries, commonly used in earlier EV designs, are particularly susceptible to both calendar and cycling-induced aging, especially at elevated temperatures and under high-state-of-charge conditions [15]. Experimental work on commercial NMC cells has shown that aging is accompanied by resistance growth, gas generation, electrode expansion, and electrolyte decomposition, all of which contribute to performance degradation and may eventually lead to failure [16].

Although capacity fade is relatively well characterized, less attention has been given to safety-related behavior of degraded batteries under high-power transient conditions, such as rapid acceleration or climbing [17,18]. Under such loads, differences in internal resistance between cells lead to uneven current distribution and increasing voltage divergence. In practice, voltage spreads exceeding approximately 200–500 mV are typically associated with abnormal operation and advanced degradation.

Cells with elevated internal resistance tend to experience disproportionately large voltage drops under load, which may push individual cells below safe operating limits. This effect is linked to increased heat generation and can accelerate further degradation or trigger internal failure mechanisms, including lithium plating or short-circuit formation [19,20]. In addition, degradation-related processes such as gas evolution, electrolyte breakdown, and mechanical deformation (e.g., swelling) are commonly identified as early indicators of thermal instability [21,22].

At the pack level, these effects are further influenced by module architecture. In configurations that include parallel cell groups, such as 4S2P layouts, weaker cells can dominate current distribution, leading to faster imbalance growth and increasing the probability of localized failure. This becomes particularly evident during high-current transients (on the order of 150–200 A), where voltage response becomes highly non-uniform.

Alongside laboratory studies, increasing attention has been given to diagnostic methods for detecting degradation in operating vehicles. Techniques such as electrochemical impedance spectroscopy (EIS), post-mortem analysis, and data-driven approaches are widely used to characterize degradation modes and estimate state-of-health (SOH) [23,24]. However, the extent to which in-vehicle diagnostics reflect actual safety limits under high-load conditions remains insufficiently understood.

In this work, this issue is examined using a real-world case of a degraded EV battery system. The study combines on-vehicle diagnostics with controlled laboratory testing of selected battery modules. Using OBD-based measurements, the full cell voltage distribution was analyzed under both low and high load, allowing identification of the most imbalanced section of the pack. These modules were then removed and subjected to laboratory testing, including capacity evaluation in a 16S2P configuration and high-current discharge experiments with thermal monitoring.

The analysis provides several key observations. First, a clear relationship is confirmed between cell imbalance detected in on-board diagnostics and the actual state-of-health measured under laboratory conditions. Second, the results show how voltage divergence develops over time under load and can lead to rapid failure of individual cells. Third, thermal and mechanical effects—such as localized heating and swelling—are shown to accompany this process. Finally, the study demonstrates that degraded battery modules may present significant safety risks during high-power operation, even when their nominal capacity remains at moderate levels.

Despite extensive research on battery aging and SOH estimation, several gaps remain when considering real-world EV operation. Most studies focus either on controlled laboratory aging or on simulation-based approaches, while fewer works analyze batteries subjected to long-term combined calendar and cycling stresses. In addition, although voltage imbalance is widely recognized as a degradation indicator, its direct link to extreme voltage collapse under high-current conditions has not been sufficiently quantified. There is also a lack of approaches that combine non-invasive vehicle diagnostics with destructive testing to validate safety limits.

For these reasons, experimental studies that connect field diagnostics with controlled high-load testing are needed. Such an approach makes it possible to better understand safety risks associated with heavily degraded battery modules under realistic operating conditions. The results can support more effective diagnostic strategies and improve decision-making in maintenance, repair, and second-life applications of EV batteries.

2. Materials and Methods

The analyzed vehicle was a 2016-model Nissan Leaf (Figure 1) equipped with a 30 kWh lithium-ion battery pack, rated at a nominal voltage of 400 V DC and a nominal capacity of 83 Ah. The traction motor was rated at 80 kW (109 hp) with a maximum torque of 254 Nm. As-found mileage at the time of testing was 106,394 km, corresponding to a relatively high real-world usage intensity for an early-generation LEAF.

The battery pack consists of multiple 4S2P modules, where each module is formed by four series-connected cells and two parallel strings, resulting in a higher total voltage while maintaining sufficient current-handling capability. The pack is monitored and controlled by an OEM Nissan BMS, which reports the State-of-Health (SOH) and various voltage and temperature parameters both in-vehicle and via the on-board diagnostics (OBD) interface.

The scheme of the work is presented in Figure 2. The study follows a structured experimental workflow for assessing the safety of degraded lithium-ion battery modules from an electric vehicle. The process begins with vehicle selection and on-board diagnostics using the LeafSpy application connected via OBD interface. Key operational parameters, including state of charge (SOC), pack voltage, current, temperature, and individual cell voltages, are recorded under both high- and low-load conditions.

Based on the collected data, cell voltage imbalance analysis is performed to identify the most degraded segments of the battery pack. Modules exhibiting the highest voltage deviation (cells 73–88) are selected for further investigation. The battery pack is then safely disassembled, and the selected modules are extracted for laboratory testing.

In the laboratory phase, the modules are reconfigured into a 16S2P setup and subjected to controlled capacity testing within the 4.1–3.1 V range using a 600 W load, while being monitored by an external BMS system. The measured capacity is used to evaluate the state of health (SOH) and compare it with the BMS-reported value.

Subsequently, the weakest cell segment (cells 81–84) is selected for high-current discharge testing to assess safety limits under extreme operating conditions. During this test, time-resolved measurements of individual cell voltages, voltage differences, and thermal behavior are recorded, including infrared thermography.

The experiment concludes with the observation of failure mechanisms such as voltage collapse, rapid increase in cell voltage imbalance, temperature rise, and physical deformation (swelling). These results form the basis for safety assessment and practical recommendations regarding early detection and replacement of degraded battery modules.

2.1. On-Vehicle Diagnostics and Fault Description

The primary objective of the on-vehicle tests was to diagnose the root cause of a sudden drop in the indicated driving range and periodic loss of power, accompanied by the activation of the so-called “turtle mode”, where the vehicle speed is limited to about 40 km/h. The vehicle owner reported that under hard acceleration the propulsion system briefly loses power and triggers a “traction system fault” warning, after which continued driving is only possible in service mode with significantly reduced motor power. The Nissan BMS recorded error code P33E6, which is associated with cell-voltage-imbalance or low-voltage conditions in the high-voltage battery pack. To quantify the state-of-charge (SOC), high-voltage (HV) pack voltage, current and temperature, the vehicle was monitored using the LeafSpy application connected via an OBDII adapter. Typical logged parameters included SOC (≈66–74%), HV pack voltage (around 350 V at the time of the test), HV current up to about 170 A under heavy load, and several temperature readings across the pack. The BMS-reported SOC stability and range display were found to be highly erratic under dynamic driving conditions, with displayed range values dropping from about 150 km to ca. 30 km during aggressive acceleration events.

2.2. Test Setup and Data Acquisition

For on-vehicle diagnostics, an OBDII-to-Bluetooth interface (ELM327-based) was used to connect the Nissan Leaf BMS to a smartphone running LeafSpy. The software recorded real-time data from the high-voltage traction battery, including cell-voltage distribution, pack current, pack voltage, SOC, pack temperature, and 12 V auxiliary battery parameters. Cellular-level data were recorded at a few seconds’ interval, enabling a snapshot of the pack state under both quiescent and high-load conditions.

Measurements were taken under two main conditions:

• High-current condition: full-throttle acceleration starting from ca. 67% SOC, held for about 6 s at maximum available current (≈165–170 A), while recording the transient voltage drop and cell-voltage spread.
• Low-current condition: low-load, steady-state driving at around 40% SOC, where the pack current was typically below 30 A, to capture the baseline cell-voltage distribution.

All raw data were exported from LeafSpy and post-processed using Python (Matplotlib, Pandas) to generate cell-voltage bar plots and statistical summaries (min/avg/max cell voltage, cell-voltage delta, temperature spread). The same data set was used to infer the overall pack degradation and to identify the segments of cells with the most critical voltage deviation.

2.3. Module Selection, Disassembly and Safety Precautions

The highest-impact cells identified from the OBDII data were those numbered 73–88, which exhibited the largest voltage deviations and lowest voltages during high-current operation. These cells were located within four 4S2P battery modules that were therefore selected for further laboratory testing. The battery pack cover was safely removed in accordance with the manufacturer’s guidelines and workshop safety protocols for high-voltage systems.

Prior to module removal, the high-voltage traction system was deactivated, the pack was discharged to a safe SOC level, and the service-mode procedure was applied to isolate the traction battery. The four selected modules were then carefully disconnected from the pack and isolated using appropriate personal protective equipment (PPE) and high-voltage safety tools. Each module was cleaned and visually inspected for any visible damage, swelling or leakage before being connected to the test bench.

2.4. Capacity Test of Degraded Modules

Each of the four selected modules was individually equipped with a custom mechanical compression fixture to ensure the same level of compression as the original battery pack, thereby maintaining proper cell-to-cell contact and avoiding mechanical stress increase during thermal expansion. The four modules were interconnected in series to form a 16S2P assembly (16 cells in series, 2 parallel strings), which corresponds to the same cell configuration as in the original pack segment.

A third-party JK-BMS 150 A (8–16S) was used to monitor and protect the test assembly. The BMS was configured to the appropriate cell count (16S) and calibrated with known reference voltages. The pack was first charged at constant current until the voltage per cell reached 4.1 V—a level at which the Nissan Leaf BMS assumes 100% SOC, remaining below the typical 4.2 V limit used in many consumer Li-ion applications.

Following the charge phase, the test pack was discharged using a 600 W battery tester operating in constant-power mode until the cell voltage dropped to 3.1 V—a level at which the original Nissan Leaf BMS assumes 0% SOC, and which is close to its lower cut-off voltage. The total coulombic capacity delivered during the 4.1–3.1 V discharge was recorded as 49.8 Ah.

During the test, the BMS-built-in capacity estimator, cycle counter and remaining-capacity display were also recorded, along with the average cell voltage (≈3.52 V) and cell-voltage difference (≈58 mV) to quantify the remaining in-pack imbalance. Temperature sensors on the BMS board (MOS temperature and two battery temperature channels) were used to ensure the pack stayed within a safe thermal envelope (ca. 23–25 °C ambient-like conditions).

2.5. High-Current Discharge Safety Test

In addition to the quasi-static capacity test, a high-current discharge test was performed on a single weak segment (cells 81–84) to assess its safety margins under extreme load. This segment was selected based on the on-vehicle OBDII data, where cell 82 consistently exhibited the lowest voltage and the largest voltage drop under high-current conditions, indicating elevated internal resistance and/or advanced degradation.

The test was conducted without the BMS protection, using only a 600 W/40 A battery tester and a manual digital multimeter for voltage and current monitoring. The segment was initially charged to a resting voltage of about 4.04–4.05 V per cell. The test began at t = 0 min, with the tester operating in constant-power mode (600 W). When the pack voltage dropped below about 15 V, the tester automatically switched to constant-current mode (40 A) to maintain the load.

Voltage measurements for each cell (81–84) were manually recorded at predefined time intervals (0, 2, 4, 28, 33, 38, 43, 46, 51, 56 min). The cell-voltage delta was computed as the difference between the maximum and minimum cell voltages within the segment at each time step. The test was terminated at 56 min upon clear signs of cell failure:

• Electrical: one cell opened the circuit (the multimeter showed OL when the probes were placed directly on the poles). After waiting, the multimeter measured 1.2 V across that series.
• Thermal: an infrared camera recorded a local temperature of 43 °C on the exterior of the failing cell, while the ambient temperature was ca. 19 °C. The cell visibly swelled and did not return to its original shape after cooling for several hours.

All high-current tests were performed inside a ventilated enclosure with appropriate fire-safety equipment and thermal monitoring to ensure safe observation of the early-stage thermal-runaway-like behavior. The collected time-series data were used to plot the individual cell voltages and the evolving cell-voltage delta, serving as a basis for the safety assessment in the Results section.

During the high-current discharge test, individual cell voltages were recorded manually at predefined time points; therefore, the measurements captured the main evolution of cell divergence but not the full transient voltage-collapse dynamics at high temporal resolution.

Thermal behavior was monitored using infrared imaging and direct observation; however, cell surface temperature and current were not recorded as a fully synchronized continuous time series during the entire test.

Because the present study was based on a single real-world battery pack and a single destructive high-current test of the weakest segment, repeated measurements were not available for statistical treatment using standard deviation or confidence intervals. Therefore, the reported values should be interpreted as direct experimental observations from a case study rather than as population-level estimates.

For clarity, the applied load conditions are also expressed in terms of approximate C-rate. During the capacity test of the 16S2P module assembly, the discharge current was about 10.8 A at a measured pack voltage of 56.26 V, corresponding to approximately 0.13 C with respect to the nominal capacity of 84 Ah. During the high-current safety test of the weakest segment, the tester operated at 600 W and subsequently in constant-current mode at 40 A, corresponding to approximately 0.48 C when referenced to the same nominal capacity basis.

The experimental setup consisted of: (i) a LeafSpy application connected via an ELM327-based OBDII Bluetooth interface for in-vehicle diagnostics; (ii) a JKBMS 150 A, 8–16S unit for laboratory monitoring and protection of the 16S2P assembly; (iii) a 600 W battery tester used in constant-power/constant-current discharge modes; (iv) a manual digital multimeter for pointwise voltage verification during the destructive test; and (v) an infrared camera for thermal observation of the tested cells.

The safety-related thresholds discussed in this study were not established as universal certification limits. Instead, they were derived empirically from the combined interpretation of (i) OEM/BMS operating windows used for the tested Nissan Leaf battery system, (ii) the observed onset of abnormal vehicle behavior under load (P33E6 fault, turtle mode, sudden range collapse), (iii) the experimentally recorded voltage divergence, voltage collapse, localized heating, and swelling during laboratory testing, and (iv) literature-reported ranges associated with abnormal degradation behavior. Therefore, these thresholds should be interpreted as practical case-based screening indicators for the investigated battery pack rather than generally applicable limits for all EV battery systems.

3. Results

Figure 3 presents the cell-voltage distribution of the 30 kWh Nissan Leaf battery pack as recorded via LeafSpy under high-current load (165.66 A) at a SOC of 74.2%. The pack shows a total cell-voltage delta of 2.323 V (2323 mV) across the 96 cells, indicating significant imbalance during the full-acceleration event. Key metrics include Ahr = 45.16, SOH = 58.82%, Hx = 25.13%, and pack voltage around 288–290 V (inferred from average cell levels near 3 V).

A snapshot, presented in Figure 4, shows the cell-voltage distribution under low-current conditions (SOC ≈ 40%, pack current ≈ 25–30 A). Here the cell-voltage spread is markedly reduced but still non-negligible, with a minimum cell voltage of about 3.554 V, an average of 3.631 V, and a maximum of 3.721 V (delta ≈ 167 mV). The pack voltage was around 348 V under these conditions, confirming that the pack still retains a relatively high nominal voltage level despite advanced degradation. The 12 V auxiliary battery voltage remained stable at about 13.04 V, indicating that the power-management system and low-voltage components are not directly responsible for the traction-system faults.

These results confirm earlier observations in the literature that cell-voltage spreads in the tens of millivolts can be considered acceptable in healthy LEAF packs, while triple-digit millivolt deviations under load are indicative of severe cell-imbalance or advanced degradation. The observed spread in the tested pack is consistent with reported symptoms of “turtle mode” activation and range-indicator instability, as the on-board BMS likely limits power once the lowest-voltage cell approaches the cut-off threshold.

3.1. Capacity Test and SOH Assessment of Degraded Modules

Table 1. BMS-reported parameters during capacity test (16S2P, cells 73–88).

Parameter	Value
Charge	ON
Discharge	ON
Balance	OFF
Main Voltage	56.26 V
Main Current	−10.8 A
Battery Power	607.3 W
Battery Capacity	66.0 Ah
Remain Capacity	23.2 Ah
Remain Battery	35%
Cycle Count	1
Cycle Capacity	88.8 Ah
Time Emerg.	0
Time Enter Sleep	86,400 s
LCD Buzzer Alarm	OFF
Ave. Cell Volt.	3.516 V
Cell Volt. Diff.	0.058 V
Balance Curr.	0.000 A
MOS Temp.	23.6 °C
Battery T1	23.6 °C
Battery T2	25.4 °C
Detail Logs Count	129
Cell Type	Li-ion
Measured Capacity Tester (4.1–3.1 V)	49.8 Ah

presents the BMS-reported parameters recorded during the capacity test of the four selected modules (cells 73–88, configured as 16S2P). The average cell voltage was 3.516 V, with a maximum cell-voltage difference of 58 mV, indicating a moderate remaining imbalance within the segment. The pack voltage was 56.26 V at the start of the test, and the pack current was −10.8 A during the constant-power (600 W) discharge phase. The BMS-estimated capacity of the segment was 66.0 Ah, with 23.2 Ah remaining at the end of the test (≈35% of the BMS capacity).

Table 2. Voltage statistics for the tested cells (N = 16).

Metric	Min	Mean	Max	Delta
Voltage_V	3.479	3.516	3.538	0.059

summarizes the cell voltage and resistance statistics from the same test. The measured capacity obtained from the external 600 W battery tester, integrated over the 4.1–3.1 V discharge window, was 49.8 Ah. Given the nominal module capacity of 84 Ah for the 30 kWh Nissan Leaf pack, the laboratory-determined SOH is:

$${SOH}_{lab}={\displaystyle \frac{49.8}{84}}\approx 59\%$$

(1)

This value is slightly higher than the on-board BMS-reported SOH of ca. 57%, consistent with studies showing BMS-derived SOH can be conservative. Cell-wire resistances were mostly in the range of 0.038–0.059 Ω, supporting moderate internal resistance growth.

The cell voltages recorded during the test showed a maximum spread of 59 mV, with the minimum individual cell voltage at 3.479 V and the maximum at 3.538 V. This moderate imbalance is consistent with the on-vehicle observations and indicates progressive cell degradation, though no immediate critical voltage deviation was present under the test conditions.

3.2. High-Current Discharge of the Weakest Segment (Cells 81–84)

Figure 5 shows the individual cell voltages of the weakest segment (cells 81–84) during a 56 min high-current discharge test at 600 W (switching to 40 A when the pack voltage dropped below about 15 V). At the beginning of the test (t = 0 min), all four cells were at resting voltages around 4.04–4.05 V, corresponding to a high SOC level within the 4.1–3.1 V window.

The most notable behavior is observed after about 38 min of continuous discharge, when the cell-voltage spread suddenly increases and cell 82 begins to deviate significantly from the others. By 43 min, the cell-voltage delta reaches about 810 mV, with cell 82 dropping to 2.25 V while the other cells remain above 3 V. This situation persisted until the 56th minute of the test, when the voltage measurement showed 1.2 V on that cell, indicating a near-catastrophic internal failure or short-circuit. (The multimeter displayed OL when the probes were placed directly on the poles, and after a waiting period measured approximately 1.2 V.)

Figure 6 presents the cell-voltage delta (maximum minus minimum cell voltage within the segment) over time. The delta remains relatively low during the first 30 min (typically below 100–150 mV), but then rises sharply after 38 min, reaching 810 mV at 43 min and exceeding 1.5 V by the end of the test. This rapid increase in voltage spread is a known precursor to unsafe operation in lithium-ion battery packs, as the weakest cell is forced to carry a disproportionate share of the voltage drop and generates excess heat.

3.3. Thermal Behavior and Safety Implications

Thermal measurements using an infrared camera (Figure 7) show that the surface temperature of the failing cell (cell 82) reached about 43 °C at the moment of failure, while the ambient temperature was ca. 19 °C. This corresponds to a temperature rise of about 24 K above ambient, which is consistent with experimental studies showing that high-current discharge rates can increase cell surface temperature to similar levels even without full-scale thermal runaway.

Visually, the cell exhibited pronounced swelling and did not return to its original shape after cooling for several hours, indicating irreversible mechanical damage and electrolyte decomposition. The temperature of the module as a whole approached ambient condition only after about 3 h, suggesting that the thermal-inertia effects are significant in such densely packed modules. These observations are consistent with early-stage electro-thermal instability and irreversible cell damage; however, no post-mortem analysis was performed to directly confirm the underlying failure mechanism.

4. Discussion

The results show a clear link between the observed P33E6 fault, “turtle mode” activation, and sudden range drops, and the presence of strong cell voltage imbalance under high-current conditions. During a short acceleration event (165–170 A), the voltage spread across the pack reached 2323 mV, which is far beyond values typically reported for aged lithium-ion systems. In most studies, imbalance under load remains within a few hundred millivolts [25]. The magnitude observed here therefore points to a near-critical state, where the weakest cells are pushed well outside their safe operating window. This explains both the activation of BMS protection and the noticeable loss of propulsion performance. Compared with literature data, the measured voltage spread represents an extreme case of cell divergence. Even moderately degraded packs usually remain within the 100–500 mV range under high load [25]. Exceeding this level by a large margin suggests that degradation is highly localized and dominated by a small number of weak cells. At the same time, the capacity test confirms that standard on-board diagnostics can still provide a reasonably accurate estimate of overall battery condition. The agreement between the measured SOH (59%) and the BMS-reported value (57%) is consistent with previous findings showing that BMS algorithms tend to be slightly conservative [26,27]. Under moderate load, the tested modules behaved in a relatively stable manner, with a voltage spread of about 58 mV. However, this apparent stability disappears under higher current. The observed voltage drop to approximately 1.2 V is particularly significant, as typical cut-off values reported in experimental studies are in the range of 2.5–3.0 V [25]. Values close to 1 V are rarely reported and usually indicate severe internal degradation or imminent failure. The high-current test of the weakest segment provides further insight into this behavior. The rapid increase in voltage imbalance—from 20 mV to over 800 mV within a short period—highlights how strongly degradation effects can accelerate under load.

This type of response is consistent with resistance-driven divergence described for aged cells [28]. The collapse of one cell from about 4.04 V to 1.2 V suggests a near-complete loss of functionality, likely linked to internal damage such as lithium depletion or the formation of an internal short circuit. However, since no EIS or post-mortem analysis was performed, the exact mechanism cannot be confirmed. Thermal observations support this interpretation. The measured surface temperature of 43 °C (about 24 K above ambient) falls within the range typically associated with early-stage thermal instability [29]. At the same time, no immediate transition to thermal runaway was observed. This behavior is in line with reports showing that degraded cells can undergo a prolonged pre-failure phase, during which heat accumulates gradually and is accompanied by gas evolution and mechanical deformation [30]. The visible swelling of the cell indicates irreversible structural damage, most likely related to electrolyte decomposition. Interestingly, the failure occurred at relatively moderate temperatures (~40–45 °C), which is well below the levels often associated with full thermal runaway (>80 °C). This underlines the importance of detecting early-stage degradation before more severe events develop. The test was stopped once clear signs of electrical failure and thermal instability appeared, in order to maintain safe conditions. As a result, it was not possible to determine whether continued loading would have led to venting or full thermal runaway, or whether the observed open-circuit behavior represents the final failure mode in this case. A key contribution of this work is the direct combination of field diagnostics with destructive laboratory testing.

Previous studies have typically focused either on controlled aging experiments [31] or on in-vehicle monitoring data [32]. Combining both approaches makes it possible to relate real operating conditions to actual failure mechanisms. The quantitative results are broadly consistent with published data, but represent a more severe case. Voltage spreads reported in the literature usually remain within 100–500 mV, whereas the present study shows values exceeding 2 V. Similarly, the SOH agreement within 2–5% matches earlier observations, and the measured temperature rise (≈24 K) falls within commonly reported ranges for high-rate discharge [29]. Taken together, these comparisons suggest that the observed behavior follows known degradation trends, but reaches a much more critical level. The results also point to limitations of pack-level monitoring. Pack voltage and SOC remained within acceptable ranges even when one of the cells approached failure. This confirms that cell-level heterogeneity plays a dominant role in aged battery systems, particularly in configurations with parallel groups [33]. In such cases, weaker cells impose additional stress on neighboring cells, which accelerates imbalance growth and increases the likelihood of localized failure. Similar effects have been reported in second-life battery studies, where cells with SOH values between 60% and 80% may still behave unpredictably under dynamic loading [34]. This indicates that capacity alone is not a sufficient indicator of safe operation. Instead, load-dependent behavior needs to be taken into account. From a practical perspective, voltage imbalance under load appears to be a more reliable indicator of safety than static SOH values. While a battery with approximately 60% SOH may still function under normal conditions, high-power demand can expose hidden weaknesses. This observation is consistent with previous studies on second-life applications, where moderate capacity loss does not necessarily guarantee safe operation under dynamic conditions [35].

These findings have direct implications for maintenance and diagnostics. OBD-based tools can be useful for identifying problematic modules, especially if measurements are taken under sufficiently high load. In addition, targeted high-current testing combined with thermal observation provides a practical way to assess safety margins and identify failure-prone cells before reuse or repurposing. Overall, the analyzed battery system shows that moderate degradation levels do not guarantee safe operation. The underlying electrochemical heterogeneity becomes critical under high-current conditions, which are not captured by standard diagnostic metrics. This highlights the need for diagnostic approaches that account for dynamic loading rather than relying solely on static indicators.

Several limitations of this study should be noted. The analysis is based on a single battery pack, which limits the general applicability of the results. The high-current test was performed on a selected segment rather than the full pack, and therefore does not fully represent pack-level interactions. In addition, advanced diagnostic techniques such as EIS or post-mortem analysis were not used, which restricts interpretation of the underlying mechanisms. Finally, the study focuses on short-term behavior and does not address long-term thermal propagation or full thermal runaway. Future work should extend this approach to larger datasets and include more detailed characterization methods.

5. Conclusions

The results obtained from this study highlight a key issue in aging EV batteries: acceptable capacity does not necessarily mean safe operation. Although the tested pack still retained roughly 60% SOH, its behavior under high current revealed clear instability. In particular, the rapid voltage collapse observed in one of the cells shows how strongly local degradation can affect the entire system. This type of behavior is not visible in standard capacity tests and may remain undetected during normal operation until a critical situation occurs. From a practical point of view, the study confirms that simple diagnostic tools such as LeafSpy can be useful for identifying problematic cells, especially when measurements are taken under load. However, relying only on static indicators like SOH can be misleading. The experimental approach used here—combining field data with targeted laboratory testing—provides a more realistic picture of battery condition. While this work is based on a single case, the observed patterns are consistent with known degradation behavior and highlight the need for more load-based diagnostic methods in aging EV fleets.