Quick answer: thermal runaway in a battery pack is mitigated, not prevented, through three layers working together: cell-level choices (format, spacing, matched cells), module and pack-level containment (barriers and venting sized to the actual mass budget), and BMS-level monitoring that cuts current and isolates faults before a marginal cell becomes a propagating one. None of these is a guarantee. Together they define how much margin a pack actually has.
An EV pack can typically spend 15-20 kg on aerogel barriers, ceramic-fiber sheets, and a liquid cold plate to slow a thermal event from spreading. A drone or robotics pack can't spend that mass budget on anything, let alone thermal materials, and still meet its endurance target. The cell chemistry and the physics of propagation are the same in both cases. The design response has to be different.
This post is about design mitigation, not elimination. No pack design "prevents" thermal runaway in an absolute sense. What a good design does is reduce the odds of a cell entering runaway, slow or contain propagation if one does, and give the system time to fail safely instead of catastrophically.
Treating thermal design as a bolt-on, an off-the-shelf BMS here, a generic "add a barrier" there, produces a pack that looks safe on a spec sheet and behaves unpredictably in the field. Thermal margin is a system-level decision, made against the real mass and volume budget of the application, not an afterthought layered on at the end.
Why EV Thermal Solutions Don't Transfer to a Weight-Constrained Pack
Most published guidance on thermal runaway prevention is written for automotive packs, and it shows. Aerogel sheets often rated to survive temperatures upward of 1,000°C, mica and ceramic-fiber barriers between every module, liquid cold plates: these are real, effective materials, sized for a pack that can absorb hundreds of kilograms (typically in the 400-600 kg range) without threatening the vehicle's performance envelope.
A UAV or robotics pack doesn't have that option. Every gram spent on a thermal barrier is a gram not spent on flight time, payload, or range. That trade-off has to be made explicitly, cell by cell and layer by layer, not defaulted to "as much barrier as the enclosure allows."
This is the frame for everything that follows: every mitigation below is evaluated against what it costs in weight, not only what it buys in safety margin. A choice that's obviously correct in a 400 kg battery pack can be the wrong call entirely at 4 kg.
What Happens in Thermal Runaway
A cell enters thermal runaway when internal heat generation outpaces heat dissipation, usually triggered by an internal short, mechanical damage, overcharge, or an external heat source. Past a threshold temperature, the cell's own chemical reactions become self-sustaining and generate more heat than any passive cooling path can remove. The cell vents gas, and in the worst case, ignites.
The risk that pack design actually controls is propagation: whether one cell's failure stays contained or triggers its neighbors. Design choices at every level either slow that chain or accelerate it. Cell spacing, vent direction, barrier placement, and BMS response time all shift the odds, but none of them changes the underlying chemistry. A pack design manages propagation risk. It doesn't eliminate the possibility of a cell failing in the first place.
Cell-Level Design Choices That Affect Thermal Risk
Cell format and chemistry trade-offs
Cylindrical and pouch cells fail differently, and that difference matters for pack design before a single barrier gets specified.
A cylindrical cell has a rigid steel can and a designed vent path (typically a scored top cap) that directs gas and pressure in a controlled direction. That predictability is useful: a pack designer can orient cells so vent gas exits away from neighboring cells and electronics. Molicel's M65A, a 21700 cylindrical cell at 322 Wh/kg, is the highest energy-density cylindrical cell in Dan-Tech's catalog. Cylindrical cells like it make up most of what Dan-Tech builds packs with.
A pouch cell has no rigid can. It swells before it vents, and when it fails, the failure mode is less directional: the pouch can rupture along any weak point in the seal, and stacked pouch cells present more surface-to-surface contact area for heat to transfer to a neighbor. Pouch cells also tend to reach higher gravimetric energy density: Amprius's SA504 reaches 383 Wh/kg, the highest energy density in Dan-Tech's catalog, but that ceiling comes at a premium price tier. Most of the catalog's standard cylindrical cells, including options from Molicel, Reliance, and Amprius's own 21700 cylindrical cells, price well below that premium pouch tier. The decision to reach for a premium pouch cell should be driven by an actual energy-density requirement, not a default.
Chemistry matters too. Amprius's SiCore silicon-anode cells and conventional NMC graphite-anode cells (Molicel, Reliance, and most of the market) have different thermal behavior under abuse conditions, largely because of how the anode material responds to heat and mechanical stress. Neither chemistry choice removes the need for spacing, barriers, and BMS protection. It changes the margin those layers are working with, which is a question to raise with cell datasheets and a pack partner directly, not something to assume from a marketing chemistry name.
Cell spacing and orientation as propagation-risk levers
Physical distance between cells does two things: it increases the thermal mass a runaway event has to heat through before reaching a neighbor, and it creates room for the gas and heat a venting cell releases to disperse instead of concentrating on the cell next to it.
Orientation matters as much as distance. A pack designed so that each cell's vent path points toward an open channel, not toward another cell's body, buys propagation-delay margin without adding a gram of barrier material. That's the cheapest mitigation available and it's purely a layout decision, made at the CAD stage, not a materials purchase.
The trade-off is volumetric: more spacing means a larger pack for the same energy content, which works against the same mass-and-volume budget that rules out thick barriers. Spacing decisions in a weight-constrained pack are usually a matter of a few millimeters, sized against thermal simulation and the specific cell's vent geometry, not a blanket rule.
Production consistency as a thermal-risk factor
A cell that's out of spec on capacity or internal resistance, or a pack where cells weren't matched before assembly, carries elevated thermal risk that has nothing to do with barrier design. An unmatched cell in a series string can be pushed outside its safe operating window by the pack's own BMS behavior, simply because it's not carrying the same load as its neighbors.
This is a factor almost no published guidance on thermal runaway addresses, because most of it is written from the materials-vendor side, where the pack itself is someone else's problem. It's also the layer a pack builder controls directly and a component vendor can't: matching cells, running incoming inspection on every batch, and catching workmanship variance before assembly, not after. Dan-Tech builds and owns pack assembly rather than handing it to a separate contract manufacturer, which is exactly where this control has to sit.
Module and Pack-Level Design Choices
Once cell-level choices are set, the next layer is mechanical: how the pack contains a fault that does occur, without spending mass it doesn't have.
Propagation barriers. Thin ceramic-fiber or mica sheets between cell groups slow heat transfer at a fraction of the mass cost of a full aerogel layer. The design question in a weight-constrained pack isn't whether to use a barrier, it's where. A barrier between every cell adds real weight for a return that fades once spacing and venting are already doing their job; a barrier at the boundaries between cell groups, where a runaway event would otherwise cross into a much larger section of the pack, buys more containment per gram.
Vent pathway and enclosure design. The pack housing should give vented gas somewhere to go that isn't back into the electronics bay or the airframe structure. That's a mechanical design decision made early, not a material added late: a housing with a dedicated vent channel needs no extra mass to do this job; one without has to make up for it with barrier material everywhere.
Structural containment. The enclosure's job in a fault event is to buy time, not to be fireproof. A structure that stays intact for the seconds it takes a BMS to cut current and isolate a fault is doing real work, even if it isn't rated to contain a fire indefinitely. That's a realistic target for a mass-constrained enclosure, and a more honest one than claiming full containment.
The organizing principle across all three: every gram of containment should be justified against what it actually buys in propagation delay, not applied uniformly because it's what an automotive reference design does.
BMS-Level Protection: What Thermal Cutoff Behavior Actually Does (and Doesn't)
A battery management system's thermal protection layer monitors cell or group temperature and voltage, and cuts current or isolates a fault when either crosses a threshold. Reliance's RS60 cylindrical cell, for example, carries a documented 80°C thermal cutoff at its rated continuous discharge current, a real number a BMS's temperature-cutoff logic can be set against.
That's meaningfully different from preventing thermal runaway. A cutoff acts on measured temperature and voltage, which means it responds after a cell has already started drifting outside its safe window, not before. It can stop a marginal cell's condition from worsening under continued load. It doesn't undo damage that's already occurred, and it can't respond to a failure mode it isn't monitoring, like mechanical puncture.
This is worth being direct about, because it's the single most common overclaim in thermal-runaway marketing content: a BMS is one protective layer, not a guarantee. A pack with a well-specified BMS and no attention to cell spacing or matching is not meaningfully safer than one with neither. The BMS's real job is fault isolation and early response, working alongside the cell-level and pack-level choices above, not replacing them.
What to Ask a Custom Battery Pack Partner About Thermal Design
Most guidance on this topic is written from the materials-vendor or contract-engineering side: here's what thermal runaway is, here's roughly how it's mitigated. As the engineer briefing a pack partner, the more useful list is what to ask before the design is locked:
- What temperature and voltage thresholds trigger the BMS's thermal cutoff, and at what response time?
- What's the rationale behind cell-to-cell spacing in this specific pack, not spacing in general?
- Where are propagation barriers placed, and what's the weight trade-off against the alternative of more spacing or better venting?
- How are cells matched before assembly, and what's the incoming QC process on cell batches?
- Can the partner show simulation or test data for this pack's specific thermal design, not just the cell's datasheet numbers?
- What does a documented fault event look like, i.e. what evidence would the partner actually be able to produce if something went wrong in the field?
A partner who can answer these with specifics, not general reassurance, is doing the work. A partner who defaults to "we use a BMS" as the whole answer isn't.
Standards That Are (and Aren't) Relevant to UAV and Robotics Packs
UN 38.3 and IEC 62133 are real, relevant standards, and Dan-Tech packs are designed to meet them where a project scopes them in. It's worth being precise about what they actually cover: UN 38.3 addresses transport safety (the conditions a cell or battery has to survive in shipping), and IEC 62133 addresses general cell and battery safety requirements. Neither is a thermal-runaway-propagation standard specifically, and neither should be cited as evidence that a pack contains or delays propagation.
Two standards that come up frequently in automotive thermal-runaway content don't apply here. UL 2580 is explicitly scoped to on-road electric vehicles and explicitly excludes light electric vehicles, which puts it outside Dan-Tech's UAV and robotics scope entirely. China's GB38031 "5-minute rule," the regulation requiring no fire or explosion for at least 5 minutes after a thermal event begins in a cell, with the passenger compartment kept smoke-free during that window so occupants have time to evacuate, is a passenger-compartment safety window for automotive cabins. It's a real, specific standard, and it has no equivalent scope in a drone or robotics application with no passenger compartment to protect. Citing either as general industry practice for a UAV or robotics pack would be borrowing a number that doesn't mean what it appears to mean.
Summary: Key Decisions
- Thermal runaway is mitigated, not prevented. Every claim in this space should say "reduces" or "delays" propagation risk, never "prevents" it.
- Weight-constrained packs can't use automotive-scale barriers and cold plates. Every mitigation choice has to be evaluated against the actual mass budget, not a car's.
- Cell format and chemistry set the starting risk profile before any barrier is added: cylindrical cells vent more predictably; pouch cells reach higher energy density at a real weight and cost trade-off.
- Cell spacing and vent orientation are free propagation-delay margin, available at the layout stage with no added material.
- Production consistency, cell matching and incoming QC, is a real thermal-risk factor that most published guidance skips entirely.
- A BMS thermal cutoff is one protective layer, not a guarantee. It responds to measured conditions; it doesn't prevent a fault from starting.
- UN 38.3 and IEC 62133 are real standards Dan-Tech packs are designed to meet where scoped. UL 2580 and China's 5-minute rule are automotive-specific and don't apply to UAV or robotics packs.
FAQ
Can a battery pack design "prevent" thermal runaway?
No pack design prevents thermal runaway in an absolute sense. Design choices at the cell, module, and BMS level reduce the likelihood of a fault and slow or contain propagation if one occurs, but none of them eliminates the underlying chemical risk.
What causes thermal runaway in a lithium-ion battery pack?
A cell enters thermal runaway when internal heat generation outpaces heat dissipation, typically triggered by an internal short, mechanical damage, overcharge, or external heat. Past a threshold temperature, the reaction becomes self-sustaining faster than any passive cooling path can remove heat.
Does a BMS guarantee a pack won't experience thermal runaway?
No. A BMS thermal cutoff monitors temperature and voltage and cuts current or isolates a fault after a threshold is crossed. It's a real protective layer, but it acts on measured conditions after drift has already started, not before a fault begins.
Why can't UAV and robotics packs use the same thermal barriers as EV packs?
EV packs can spend tens of kilograms on aerogel sheets and liquid cold plates without threatening the vehicle's performance. A UAV or robotics pack has no equivalent mass budget, so every gram spent on a barrier is a gram not spent on flight time, payload, or range, and mitigation has to be chosen accordingly.
Do UN 38.3 and IEC 62133 cover thermal runaway propagation?
No. UN 38.3 covers transport safety and IEC 62133 covers general cell and battery safety. Neither is a propagation-specific standard, and automotive standards like UL 2580 or China's 5-minute rule don't apply to UAV or robotics packs at all.
Thermal design isn't something a custom pack partner should promise in a blog post. It's something worth walking through together against your actual application, mass budget, and duty cycle. Start a conversation about your pack's thermal design in the ToolBox, or browse the underlying cell options in Dan-Tech's Lithium Ion Battery Packs catalog.
.png){kind=small}
