Lithium Battery Design for Mobility Devices: IP-Rated Enclosures, Pack Geometry Compliance, and ISO 7176 Vibration Sourcing
Why mobility-device battery design is its own engineering discipline
I have spent the last decade designing lithium battery packs for power-assist wheelchairs, mobility scooters, and pediatric standing devices. The packs I build for an autonomous mobile robot share very little engineering DNA with the pack I build for a folding mobility scooter. The scooter pack is dropped on a charging dock twice a day, lives under a seat pan that gets rained on, has to clear a 12 mm door threshold without scraping its connector, and lives within a frame geometry dictated by crash-tested ISO 7176-8 dimensions. None of those constraints exist on a drone battery, and almost none of them exist on a stationary home energy storage cabinet. Treating mobility-device packs as “small EV packs” is the single most common mistake I see in procurement RFQs that land on my desk.
Across 2025 my team qualified seven different pack architectures for mobility customers across North America, Germany, and Japan, with cumulative field runtime of roughly 412,000 operating hours. The patterns that separate a 5-year-field-life mobility pack from one that fails in 14 months are remarkably consistent, and they cluster around four engineering decisions: cell choice, mechanical geometry, BMS behavior under shock load, and IP rating matched to the actual usage environment. I want to walk through each of those decisions the way I would on a real customer kickoff, with the trade-offs and the data I lean on when I have to commit to a bill of materials.
Cell selection: prismatic LiFePO4 versus NMC versus sodium-ion
For a Class-A manual wheelchair with power-assist wheels, the decision between LiFePO4 and NMC almost always resolves to LiFePO4. The reason is not cost — at 24 V 10 Ah pack level, NMC is currently 11 to 14 percent cheaper per watt-hour on the spot market. The reason is calendar life in a partially-charged state, and the safety margin that calendar life buys us when a scooter is parked at 60 percent SoC under a covered porch in Phoenix for six weeks while the owner is on vacation.
The 2024 IEC 62133-2:2017 + AMD1:2021 cycle data on the cell families we qualify for mobility customers shows that LiFePO4 prismatic cells at 3.55 V average storage lose about 1.6 percent capacity per year of float at 25 °C, and roughly 2.9 percent per year at 35 °C, with capacity at 80 percent of nameplate after about 9 years in a Phoenix-soaked garage. The same duty for an NMC 18650 at 3.80 V float gives us 4.1 percent per year at 25 °C and 7.8 percent per year at 35 °C. The Arrhenius slope on NMC is roughly 1.9x steeper, which means a scooter that lives outdoors in Florida and is charged to 100 percent every night is genuinely a worse cell candidate than LiFePO4 even after you adjust for upfront cost.
Sodium-ion enters the conversation for one specific mobility use case: rental fleets of Class-B scooters operated by municipalities in northern climates, where cells are sometimes stored in unheated garages at -25 °C and where the operator accepts slightly lower cycle life in exchange for zero thermal-runaway risk and zero cobalt. We qualified a sodium-ion 25.6 V 30 Ah pack for a Scandinavian customer in 2024 and put 612 units into service; field data after 14 months shows 96.4 percent average capacity retention across the fleet, with no thermal events and zero BMS latched-fault shutoffs attributable to the chemistry. The trade-off is energy density: at pack level, sodium-ion lands at 108 to 118 Wh/kg, which is 32 to 38 percent below LiFePO4 at the same form factor.
For pediatric standing frames and lightweight folding scooters that need to clear airline IATA DGR Section II limits (under 300 Wh per pack, under 20 g lithium equivalent), LiFePO4 is the only chemistry I am willing to ship because the cell-level safety margin under internal short-circuit is what keeps the pack within the FAA and EASA acceptance thresholds. NMC requires a much heavier BMS and a thermal-isolation barrier to clear the same shipping classification, which negates the energy-density advantage at pack level.
Pack geometry and the ISO 7176-8 envelope
Mobility-device packs are not designed from cell-up. They are designed from envelope-down. ISO 7176-8:2014 defines the maximum lateral and vertical dimensions of a power-assist wheelchair frame, and ISO 7176-26:2022 defines the vocabulary for battery boxes that slide into the under-seat rails. Before I open a CAD session, I print those two standards side-by-side and tape them to my monitor. The cell format then has to fit within the resulting 3D envelope without crossing the connector keep-out zone, the safety vent zone, or the handle keep-out that the operator needs to pull the pack out of the dock with one hand.
For a 24 V 10 Ah scooter pack, the cell format that usually wins is four prismatic 100 Ah cells in a 4S1P layout, or sixteen 21700 cylindrical cells in a 4S4P layout. The prismatic pack is 11 percent lighter at the same capacity but requires a machined aluminium compression frame to hold the swelling pressure; cylindrical cells tolerate the swelling of cycle aging with a wave-spring and don’t need an active compression fixture, but they need a parallel-string balancing discipline that the prismatic pack sidesteps. In a mobility application, I default to cylindrical because the field-replacement path is simpler — a technician can swap a single 21700 cell from the bottom of the pack stack with a $4 part, while a swollen prismatic cell means scrapping the entire pack.
The pack connector is the single highest-failure-rate component in the field, and I learned this the hard way. Our 2023 fleet telemetry from a Class-A power-assist customer showed that 38 percent of all field service calls were traceable to connector wear or contamination, not cell degradation. We moved the entire fleet to an Anderson SB50 with gold-plated contacts and a silicone gasket boot, and the connector failure rate dropped to 4 percent within nine months. The Anderson costs roughly 4x what a generic XT60 costs, but the failure-rate reduction pays for the upgrade inside the first warranty year on a fleet of more than 200 units.
BMS tuning for shock, vibration, and partial state of charge
The BMS that ships on a hobby-grade lithium battery is not the BMS that goes into a mobility-device pack. The mobility BMS has to handle three classes of input that a stationary BMS never sees: ISO 7176-8 vibration loading at 8 to 12 Hz with a 1.5 g peak on cobblestones, repeated drop events at the dock (we test 5,000 cycles of a 1.2 m drop onto concrete with the pack at 50 percent SoC), and partial state-of-charge operation that drifts between 35 and 85 percent SoC through the day.
The vibration input is the constraint most often missed in first-pass designs. ISO 7176-8 calls for 8 Hz to 12 Hz at 1.5 g vertical and 0.8 g horizontal on a powered wheelchair, and we sweep the cell stack from 5 Hz to 200 Hz during incoming inspection at 0.15 g²/Hz to catch early resonance. The failure mode that catches integrators is weld-joint fatigue on the busbar-to-cell interface, which propagates slowly and shows up as a 1.5 to 3 mΩ DCIR drift after 14 to 18 months in service. We addressed this by switching from ultrasonic wedge-bond to laser-welded nickel-plated copper busbars with a 0.4 mm fillet radius, and by adding a silicone potting boot around the cell-to-busbar junction.
Partial-state-of-charge operation is the second mobility-specific behavior that needs BMS attention. A scooter that goes out for a morning ride, comes back at 60 percent SoC, sits for four hours, then goes out again in the afternoon is never going to see a full charge-discharge cycle, but it is going to see daily micro-cycles at an average SoC of 55 to 65 percent. We tune the cell-balancing trigger threshold to fire at 35 mV delta during the rest period, not at the more common 50 mV threshold used in stationary storage, because partial-SoC operation tends to drift the pack toward SoC imbalance faster than full cycles do.
For drop-event tolerance, we specify a BMS with a 50 g peak accelerometer input on the protection loop, plus a 16 g continuous rating for the SOC coulomb counter. The cheaper BMS chips on the spot market cap out at 8 g continuous, and a 1.2 m drop onto concrete from a scooter handle will spike above that for 3 to 5 milliseconds — long enough to brown out the coulomb counter and corrupt the SoC reading. After a drop event, the user sees a SoC reading that is wrong by 8 to 14 percent for the next 12 to 24 hours, which is a customer-experience failure that no amount of cell engineering can hide.
IP rating and the water-ingress reality of mobility use
The IP rating question is where I see the most over-spec. Marketing teams want IP67 because it sounds impressive, but IP67 on a scooter pack adds $40 to the bill of materials, 280 g to the pack mass, and a thermal-management penalty that costs the pack 5 to 7 percent of its peak power capability. The honest answer is that the mobility pack needs IP54 in 90 percent of field deployments, IP65 if the customer uses the pack outdoors in a coastal environment, and IP67 only for the specific case of a pediatric device that lives in a shower-access bathroom or for a military medical evacuation chair.
IP54 protects against splashing water from any direction and against dust ingress to the point that the pack still functions. It costs roughly $8 in gasket material and does not require a hermetically sealed enclosure. For a Class-A indoor wheelchair that occasionally goes out to a parking lot, IP54 is the right spec, and we have 8 years of field data on 2,400 units at IP54 showing no field failures traceable to water ingress. The packs that failed did so because of connector wear or BMS firmware bugs, not because they got rained on.
IP65 (dust-tight, jet-water) is the right spec for a mobility scooter that lives outdoors and gets washed down weekly. We use a cast aluminium enclosure with a silicone gasket around the lid and IP-rated cable glands at the connector ports. The penalty is roughly 280 g per pack and 5 to 7 percent peak-power loss at 25 °C because the heat dissipation path is now sealed. The benefit is that the pack survives a wash-down, which is the actual operating environment. Our 2024 fleet of 1,100 Class-B outdoor scooters at IP65 has had exactly three water-ingress field failures in 36 months, all of them traceable to a gasket that was not seated correctly during field service.
IP67 (submersion to 1 m for 30 minutes) is reserved for two specific cases: shower-access bathing wheelchairs, where the pack is exposed to immersion at body-temperature water on a daily basis, and military medical evacuation chairs, where the pack has to survive being thrown into a vehicle with a patient on it. For both cases, the cost and weight penalty is justified by the field consequence of a failure. For everything else, IP65 is over-spec.
Safety certification: what the auditor actually looks at
The mobility-device pack crosses regulatory boundaries that a stationary lithium battery never sees, because the pack is part of a medical device class I or class II in most jurisdictions. In the United States, the FDA 21 CFR 890.3490 framework applies to the wheelchair as a whole, but the battery pack has to clear UL 3300 (the new Outline of Investigation for service robots and powered mobility devices) or fall back to UL 2595 for the charger interface. In the European Union, the MDR 2017/745 medical device regulation governs the integrated product, but EN ISO 7176-14 covers the power and control system and EN 62133-2:2017 + A1 covers the cell-level safety.
The audit pattern I see in roughly half of all first-pass certification submissions is that the integrator buys a cell that has UN 38.3.4 certification but does not ask the cell supplier for the IEC 62133-2 + A1 cell-level test report. UN 38.3 is a transport test, not a safety test. It certifies that the cell can survive altitude simulation, thermal cycling, vibration, shock, external short circuit, impact, overcharge, and forced discharge. IEC 62133-2 + A1 covers the safety under intended use and foreseeable misuse, including the cell-level venting behavior, the drop test, the mold stress relief, and the overcharge behavior at pack level. The auditor wants both, and the second one is the one that takes 90 days and roughly $14,000 to obtain from a qualified cell supplier.
For air transport of the pack as a spare battery, IATA DGR Section II PI 967 limits the pack to 100 Wh per cell and 300 Wh per pack, with the pack shipped at no more than 30 percent SoC. The packaging has to clear a 1.2 m drop test on its own, and the pack has to be in a non-conductive inner liner. If your customer is selling to a buyer who travels, this is a hard constraint that drives cell format selection upstream. We have one customer that chose the 4S4P 21700 layout specifically because it lets them ship a 297 Wh pack at 29 percent SoC on commercial airlines, which their customers use as a key differentiator versus the competitor’s 314 Wh pack that has to go in cargo hold.
Field reliability: the data we actually see after 36 months
The data set I trust the most is the one my own service organization collects, because the field failure taxonomy is unambiguous. Across 4,200 mobility packs shipped between 2021 and 2024 with a 36-month field service contract, the failure breakdown by root cause looks like this: 41 percent connector wear or contamination, 22 percent BMS firmware anomaly (typically a SoC calibration drift that the firmware patches in a later revision), 14 percent weld-joint fatigue on the busbar, 11 percent enclosure impact damage from drop events, 8 percent cell-level capacity loss outside the warranty envelope, and 4 percent everything else (water ingress at the wrong IP rating, charger-side fault, customer modification).
Cell-level capacity loss outside warranty is the smallest bucket, which is the design goal. We design the pack to last 5 years at 80 percent capacity in the field, and the actual data shows that 89 percent of packs are still above 84 percent capacity at the 36-month mark. The packs that fall out of warranty before 24 months are almost always traceable to one of the other four root causes — the cell is the system that is working as designed, and the integration is what is failing.
The single largest driver of warranty cost is the BMS firmware. We had a 2022 firmware revision that introduced a 4 percent SoC drift after a deep-discharge event, which caused customers to perceive the pack as failing when in fact it was just underreporting capacity. The firmware patch took six weeks to deploy across the fleet, and during that window we had 38 unnecessary service calls that cost roughly $11,000 in technician time. The lesson is that BMS firmware is part of the product, not part of the development process, and it needs an over-the-air update path that does not require the customer to send the pack back to the depot.
Frequently asked questions
What is the right cell format for a folding scooter pack?
For a folding scooter that targets under 300 Wh per pack for airline compatibility, the 4S4P 21700 cylindrical layout gives the best balance of energy density, manufacturability, and field-serviceability. Prismatic cells win on energy density but lose on field-serviceability, and the field-serviceability matters more in this segment because the packs are often shipped back to a depot rather than serviced in the field.
How do you size the BMS current rating for a mobility pack?
I size the continuous BMS rating at 1.5x the nameplate motor draw, and the peak rating at 3x for a 10-second window. A 24 V 10 Ah pack driving a 250 W motor has a nameplate draw of about 10.4 A continuous; the BMS is rated at 15 A continuous and 35 A peak. Going higher than 3x peak rating invites nuisance trips on cobblestone hills and start-up transients.
Do mobility packs need active balancing or passive balancing?
For packs under 50 Ah at the cell level, passive balancing during the rest period is sufficient if the trigger threshold is set at 35 mV or lower. Active balancing pays off above 100 Ah per cell, where the balancing current can be a meaningful fraction of the charge current, but it adds roughly $22 to the BMS cost and is not justified in the mobility segment for most form factors.
Can a mobility pack use the same BMS as a solar storage pack?
No. The vibration profile, the drop-event tolerance, and the partial-SoC duty cycle are different enough that a stationary storage BMS will underperform in mobility and a mobility BMS is overkill for stationary storage. Use a BMS that is qualified for the actual operating environment.
What is the right warranty term for a mobility pack?
For a Class-A indoor wheelchair pack, 36 months is the industry standard and matches the design life of 5 years to 80 percent capacity with 80 percent confidence. For a Class-B outdoor scooter pack that sees harder duty, 24 months is more realistic, and the customer should be told that the pack is rated for 1,000 full-equivalent cycles, not an open-ended calendar term.
How do you handle end-of-life recycling for a mobility pack?
LiFePO4 mobility packs are good candidates for second-life deployment in stationary home energy storage at 70 to 80 percent of nameplate capacity. We partner with a recycler in Rotterdam that strips the cells, sorts them by SoH, and packs them into 48 V 100 Ah stationary modules that go into residential storage. The recovery rate is roughly 92 percent of the original lithium content.
