Li-ion Battery-Pack Thermal Logger for Cell Temperature Gradients vs C-Rate

Li-ion Battery-Pack Thermal Logger is an Automobile Engineering test rig born from the question India asked in 2022, when a series of electric two-wheeler battery fires forced a tightening of AIS-156: what actually happens to cell temperatures inside a small pack as you push the charge and discharge rate?

Our rig is a 3S 18650 Li-ion pack (three cells, BMS-protected, charged by a hobby balance charger) with a DS18B20 digital sensor bonded to each cell, one on the ambient reference and one on a bolted aluminium heat spreader we add for the second test series. An INA219 logs pack current and voltage, a constant-current electronic load module discharges the pack at controlled C-rates, and an Arduino Uno samples everything at 1 Hz, streaming timestamped CSV to a laptop.

The experiment is deliberately simple and repeatable: charge and discharge at 0.5C, 1C and 1.5C, first with bare cells, then with the spreader fitted, and compare peak temperature rise, cell-to-cell gradient and cooling time constants. The Python pipeline turns each run into curves and a datasheet-limit comparison table. Total cost is about ₹4,300, every run is logged, the firmware enforces a 55 °C cut-off, and the report connects our small-pack gradients to the pack-level thermal-propagation thinking behind AIS-156 — an experiment an examiner can replay number by number.

Electric two-wheeler and three-wheeler adoption in India is growing fast, and so is scrutiny of battery safety: the 2022 fire incidents and the resulting AIS-156 amendments made thermal behaviour a regulatory subject, not a datasheet footnote. Yet in college laboratories battery testing usually means reading pack voltage and current; per-cell temperatures are never logged, so students never see the cell-to-cell gradient that develops with rate and imbalance — the very behaviour that pack-level safety standards regulate.

Commercial thermal chambers and cyclers that can show this cost lakhs. There is a need for a low-cost, per-cell thermal logging rig that charges and discharges a small Li-ion pack at controlled C-rates, records every cell's temperature, current and voltage at 1 Hz, enforces hard safety cut-offs, and yields quantitative results — peak temperature rise versus C-rate, cell-to-cell gradient, and the effect of a passive heat spreader — with every point traceable to a logged file and comparable against the cell datasheet's limits.

Abstract

Battery thermal behaviour decides EV safety and life, yet it is invisible in most college laboratories. This project builds a small instrumented Li-ion pack — three 18650 cells with a BMS — in which every cell carries a DS18B20 temperature sensor, an INA219 logs pack current and voltage, and an Arduino streams one CSV line per second through charge and discharge cycles at 0.5C, 1C and 1.5C. A second test series repeats the cycles with an aluminium heat spreader bonded across the cells. The analysis quantifies peak temperature rise versus C-rate, cell-to-cell gradient, and cooling time constants with and without the spreader, and compares all results against the cell datasheet's operating limits.

Introduction

After the electric two-wheeler fire incidents of 2022, the Ministry of Road Transport amended AIS-156 to mandate stricter cell, BMS and pack-level requirements, including thermal-propagation assessment. For a student, the standard's wording is abstract; our rig makes the underlying physics touchable. We chose a 3S pack of 2.5 Ah cells — about 28 Wh, roughly a hundredth of a scooter battery — so every question is safe to ask.

Existing system and gap

• Commercial battery cyclers and thermal chambers: cost lakhs and are rarely available to undergraduate teams.
• Typical student EV builds: measure pack voltage and current, almost never per-cell temperature, so gradients — the early signature of imbalance and trouble — go unrecorded.
• Teaching demonstrations: usually plot datasheet curves rather than measured data.
• Gap: an affordable, logged, per-cell thermal experiment that shows how C-rate and a simple passive heat spreader change a pack's temperature behaviour.

Proposed system

• Test article: 3S1P pack of 18650 cells in printed holders with a 3S BMS, fused leads, and a fire-safe tile working surface.
• Sensing: one DS18B20 per cell plus ambient and spreader-surface sensors on one 1-Wire bus; INA219 high-side monitor for pack current, voltage and power.
• Load and charge: a constant-current electronic load for repeatable 0.5C/1C/1.5C discharges; the hobby balance charger for standardised charges.
• Control and logging: Arduino firmware samples at 1 Hz with a 55 °C abort and streams timestamped CSV; a Python pipeline computes ΔT curves, gradients, time constants and the datasheet table.
• Experiment: paired series — bare pack versus spreader-fitted pack — with repeat runs on separate days.

Feasibility

• Technical: 1-Wire and I2C sensing, CSV logging and simple curve fitting are all second-year-level skills; the electronic load removes the need to design a constant-current sink.
• Economic: indicative cost ≈ ₹4,300 (three cells ₹900, BMS ₹150, balance charger ₹500, five DS18B20 ₹250, INA219 ₹120, electronic load ₹650, holders, wiring, fuse and tile ₹700, Arduino ₹600, enclosure and sundries ₹530).
• Operational: a full discharge run at 1C takes about an hour; two runs fit in an afternoon, and the rig lives on a lab shelf between runs.
• Schedule: bench build and verification in weeks 1–6 (VTU Phase-I), bare-pack series weeks 7–9, spreader series weeks 10–12, analysis and documentation weeks 13–15.

In scope

• One 3S1P pack, three 2.5 Ah 18650 cells, about 28 Wh, in printed holders with a BMS.
• Discharge at 0.5C, 1C and 1.5C via a constant-current electronic load; charges at the charger's standard rate.
• Per-cell plus ambient and spreader-surface temperature logging at 1 Hz; pack current, voltage and power logging.
• Bare-pack versus heat-spreader comparison, three C-rates, repeat runs on separate days.
• Analysis: ΔT curves, peak-rise versus C-rate table, gradient chart, first-order cooling time constants, datasheet-limit comparison.

Out of scope

• Thermal-runaway testing and abuse tests of any kind — the rig operates strictly within datasheet limits and aborts at 55 °C.
• Cycle-life ageing studies (hundreds of cycles) and calendar ageing.
• Active liquid or forced-air cooling; only the passive spreader is tested.
• Cell-internal instrumentation and infrared thermography (discussed as future scope).

We follow a build–verify–measure–compare plan mapped to the VTU two-phase structure: Phase-I (BXX685) covers the survey, design and bench verification; Phase-II (BXX786) covers the test campaigns and report.

Test plan: (1) verification — all five temperature channels within ±0.5 °C of the reference at ice, ambient and warm-water points; current reading within 2 per cent of the multimeter; (2) functional — the 55 °C abort verified by a warm-water test with the load disconnected from the pack; (3) campaign — every C-rate run twice, on different days, with ambient recorded so ΔT isolates the pack's own rise; (4) data — every figure in the report traceable to a named CSV file. Each objective maps to at least one test case recorded with input, expected, observed and status.

The rig is a three-layer measurement loop around the pack: sensing, guarding, and logging.

1. Test article layer: three 18650 cells in printed holders form the 3S pack; the BMS protects against over-charge, over-discharge and over-current; a blade fuse and silicone leads complete the power path; everything sits on a ceramic tile with the sand tray nearby.
2. Sensing layer: five DS18B20 sensors (three cells, ambient, spreader surface) share one 1-Wire bus, each with its unique address logged at startup; an INA219 sits high-side on the pack positive, reporting current, voltage and power over I2C.
3. Control and logging layer: the Arduino Uno polls the bus at 1 Hz, enforces the 55 °C abort by signalling the load's enable line, and streams one CSV line per second — timestamp, five temperatures, current, voltage, power, state flag — over USB serial. The host logger writes dated files; the Python pipeline computes results.

flowchart TD
  CELLS["3S 18650 pack (BMS-protected, fused)"] --> LOAD["Constant-current electronic load: 0.5C / 1C / 1.5C"]
  CELLS -->|"pack current, voltage, power"| INA["INA219 high-side monitor"]
  T1["DS18B20 on cell 1"] --> BUS["1-Wire bus"]
  T2["DS18B20 on cell 2"] --> BUS
  T3["DS18B20 on cell 3"] --> BUS
  T4["DS18B20 ambient"] --> BUS
  T5["DS18B20 on spreader surface"] --> BUS
  BUS --> UNO["Arduino Uno: 1 Hz sampling, 55 C abort, CSV stream"]
  INA -->|"I2C"| UNO
  UNO -->|"enable/disable signal"| LOAD
  UNO -->|"USB serial CSV at 1 Hz"| PC["Laptop: logger writes dated CSV files"]
  PC --> AN["Python analysis: dT curves, gradients, time constants, datasheet table"]
  CHG["Hobby balance charger (separate station)"] -->|"standardised charge"| CELLS

flowchart TD
  A["Start logging, read all sensor addresses"] --> B["Sample temperatures + INA219 every 1 s"]
  B --> C{"Any cell above 55 C?"}
  C -->|"Yes"| D["Disable load, raise ABORT flag, keep logging"]
  C -->|"No"| E{"User stops or load completes?"}
  E -->|"No"| B
  E -->|"Yes"| F["Stop, close CSV file, summarise run"]
  D --> F

Key design numbers we defend in the viva: sensor rate 1 Hz against thermal time constants of minutes, so resolution is ample; 0.5C/1C/1.5C on a 2.5 Ah cell mean 1.25/2.5/3.75 A; the 55 °C abort sits far below any danger regime while still above the highest temperature our runs should reach; DS18B20 tolerance ±0.5 °C is carried into every result's error bar.

• Forced-air comparison with a small fan on a ducted shroud, quantifying convective improvement over the passive spreader on the same matrix.
• Infrared thermography of the pack surface during runs, mapping gradients spatially rather than at three points.
• Longer-term studies: a cycle-ageing mini-campaign tracking how capacity and internal resistance drift over 50 cycles, with the same logging rig.
• Cell-voltage-per-channel logging through a dedicated multi-cell monitor so imbalance can be correlated with temperature gradients directly.
• LFP comparison series on a 4S LiFePO4 pack, contrasting the flatter voltage curve and milder thermal response with our NMC results.
• A repeatable lab experiment: PCB for the sensor bus, a written three-hour procedure and kit list so future batches extend the dataset.