18650 Battery Pack Welds Failing? The Physics Behind Reliable Spot Welds

Your battery pack has a problem. It dies after two weeks -- not the cells themselves, which test fine on the analyzer. The issue is the welds. One by one, the nickel strip connections peel away from the cell terminals, and the pack's internal resistance climbs until the BMS shuts everything down. You rebuild it. Same result.

This trouble repeats in workshops and garages everywhere, and the culprit is almost never the equipment. It is a misunderstanding of what actually happens during those few milliseconds of welding. The gap between a joint that holds for years and one that fails in days comes down to physics that most builders never learn -- thermal gradients at the weld interface, the relationship between pulse energy and nugget formation, and why the material you choose for your busbar determines everything about how you need to set your welder.

The Heat Problem Nobody Explains

Every spot weld follows Joule's Law: the heat generated at the interface equals the square of the current multiplied by the resistance and the time ($H = I^2Rt$). That squared term on the current is the entire story. Double your current, and you generate four times the heat. This is why capacitor discharge welders, which can dump thousands of amps in a single pulse, achieve results that conventional AC welders struggle to match despite drawing far more total power from the wall.

But here is the part most tutorials skip: the resistance $R$ in that equation is not constant. It is the contact resistance between the nickel strip and the battery terminal, and it changes dramatically based on surface preparation. A fingerprint on the terminal, a thin oxide layer, or a slightly warped strip that only contacts at one edge -- each of these alters the resistance, which means the same welder setting produces a different amount of heat on every single weld. This is why two identical-looking welds can have completely different mechanical strength.

The practical consequence is that parameter tuning is not a one-time setup. Before welding a production pack, you need to establish a test protocol: take three sacrificial cells, weld strips to them at your chosen setting, then peel the strips off with pliers. If the strip tears and leaves material bonded to the cell terminal, your parameters are in the right zone. If the strip separates cleanly at the interface, you are under-welding. If you see burn marks, spatter, or a hole in the strip, you are over-welding. This three-cell test takes five minutes and prevents hours of rework.

Why Nickel Is Forgiving and Copper Is Not

Nickel strips dominate battery pack construction for good reason. Pure nickel has moderate electrical conductivity (about 14% of copper by cross-section), good corrosion resistance, and -- critically for welding -- relatively high specific resistance. That higher resistance means more heat is generated right at the weld interface where you want it, making the process forgiving across a wide range of parameters.

For 0.1mm nickel, start around 15-20t on a typical capacitor discharge unit and adjust upward if you see cold welds. For 0.2mm nickel, the most commonly used thickness, the 25-35t range usually hits the sweet spot. For 0.3mm nickel, push to 40-50t. Each increment roughly corresponds to about 0.2ms of additional pulse duration, giving you a direct mental model: thicker material needs more time for heat to penetrate, but not so much time that heat spreads into the battery terminal below.

Copper tells a different story entirely. Its electrical conductivity is roughly six times that of nickel. Its thermal conductivity is about four times higher. During a weld pulse, heat generated at the copper interface conducts away from the joint almost as fast as it accumulates. The result: the material never reaches melting temperature at the interface unless you throw significantly more energy at it in a shorter window. This is why welding 0.1-0.2mm copper strips typically requires settings in the 60-80t range and, importantly, the use of flux.

Flux serves a dual purpose here. It chemically cleans the oxide layer from the copper surface immediately before welding, ensuring consistent contact resistance. It also creates a localized thermal barrier that slows heat dissipation just enough for the nugget to form. Without flux, copper welding on battery cells produces joints that look acceptable but have high contact resistance and poor mechanical strength -- they will fail under vibration or thermal cycling.

Cell Matching: The Hidden Variable

Even perfect welds cannot save a pack built from mismatched cells. When cells are connected in parallel, they self-balance to the same voltage, but current flow between them is governed by their internal resistance. If one cell has 30mΩ internal resistance and its neighbor reads 45mΩ, the lower-resistance cell will shoulder a disproportionate share of the load during discharge. Over hundreds of cycles, it degrades faster, its resistance increases further, and the imbalance accelerates until the pack fails prematurely.

The engineering standard for matching is straightforward but requires the right tools. Cells in the same parallel group should have internal resistance within 5mΩ of each other and capacity within 50mAh. You need a dedicated battery analyzer -- not just a multimeter -- because internal resistance measurement requires an AC impedance test at 1kHz, which a standard DC resistance check cannot provide.

Sort your cells into groups before you start building. Measure every cell's internal resistance and capacity, then group them so that the tightest-matched cells share parallel connections. This adds 30-60 minutes to your build time but extends pack life by 20-40% according to cycle-life testing data from cell manufacturers. It is the single highest-return investment of time in the entire process.

The Welding Pen: Your Last Mile Problem

The energy stored in the capacitor bank means nothing if it dissipates before reaching the weld. The welding pen and its cables represent the final transmission path, and their resistance directly determines how much of that stored energy actually arrives at the workpiece. A pen with 2mΩ internal resistance wastes roughly four times more energy to cable heating than one with 0.45mΩ.

This matters because cable heating is invisible. You will not see it. You will not feel it. But your weld quality will be inconsistent because as the cables warm up during a long welding session, their resistance increases, which means later welds receive less energy than earlier ones even though you changed nothing on the control panel. The practical fix: use welding pens with the largest practical copper cross-section inside (35mm² is a solid benchmark), keep the cables as short as your workspace allows, and during long sessions, pause every 50-100 welds to let the cables cool.

The welding pen tips also deserve attention. After extended use, oxidation builds up on the copper-alloy tips, creating a high-resistance layer between the tip and the nickel strip. This acts like an unintentional series resistor, stealing energy from the weld. Lightly sand the tips with fine emery cloth and clean with isopropyl alcohol before each build session. It takes 30 seconds and eliminates one of the most common sources of inconsistent weld quality.

Reading the Heat-Affected Zone

Every weld creates a heat-affected zone -- the region surrounding the weld nugget where temperatures rose high enough to alter the material's microstructure without actually melting it. In structural steel welding, a large HAZ is manageable because the base material can tolerate significant thermal exposure. Battery cells cannot.

The separator film inside a lithium-ion cell begins to lose mechanical integrity around 80-90 degrees Celsius. The electrolyte can begin to decompose at sustained temperatures above 60 degrees. These are not theoretical limits -- they represent the boundary between a cell that delivers its rated capacity for 500 cycles and one that degrades to 80% capacity in 200 cycles because repeated thermal stress during welding weakened the separator.

Millisecond-scale pulse welding restricts the HAZ to approximately 0.5-1.0mm from the weld nugget on 0.2mm nickel strip. This is the fundamental engineering advantage of capacitor discharge technology over slower welding methods. The pulse delivers energy faster than it can conduct away from the joint, creating intense localized heating at the interface while the surrounding material -- including the cell terminal below -- remains relatively cool. Research on resistance spot welding of cylindrical lithium cells confirms that properly executed pulse welds do not measurably affect cell capacity or internal resistance when measured at standard discharge rates.

The practical implication: if your welds are leaving visible heat discoloration on the cell terminal beyond 2mm from the weld point, or if you can feel warmth in the cell body immediately after welding, your pulse duration is too long. Reduce the time parameter and re-test on sacrificial cells.

Building the Pack: A Parameter Framework

Rather than prescribing fixed settings -- which would be misleading given the variation in materials, equipment condition, and ambient temperature -- here is a decision framework for parameter selection.

Start by identifying your interconnect material and thickness. For pure nickel strips at 0.1-0.15mm, begin testing at the lower end of the energy range. For 0.2mm nickel, start in the middle. For copper at any thickness, start near the maximum and plan to use flux. Always test on sacrificial cells first.

Next, consider your cell arrangement. Tight-packed configurations (cells touching) require shorter welding pens and more careful angular control to avoid the tip contacting the cell wall. Spaced configurations (2-3mm gap) provide better airflow and easier weld access but reduce energy density per unit volume. Choose based on your application's thermal and spatial constraints.

Finally, sequence your welding operations strategically. Perform all high-energy welds first -- busbar connections, thick nickel interconnects -- before installing the BMS or any sensitive electronics. High-energy pulses can induce voltage spikes in nearby conductors, and while the risk is low, the cost of a damaged BMS is high. After the BMS is installed, use reduced energy settings for the sense wire connections.

The verification protocol is non-negotiable. After completing the pack, measure the internal resistance of every weld joint using a micro-ohmmeter. The resistance of a good weld should be under 1mΩ per joint. Any joint measuring above 2mΩ should be re-welded. Then run a full charge-discharge cycle while monitoring individual cell voltages. Voltage divergence greater than 50mV between parallel-connected cells indicates a poor connection or a mismatched cell that slipped through sorting.

The Unseen Cost of Skipping Steps

Every shortcut in battery pack construction has a failure mode, and those failures do not announce themselves immediately. A weak weld might pass a bench test and survive for three months before vibration from normal use fatigues the joint. A mismatched cell might deliver acceptable performance for 100 cycles before the imbalance becomes critical. Skipping the verification protocol saves an hour today but creates a failure that takes a weekend to diagnose and repair.

The physics of battery pack welding is not complicated. It is exacting. The difference between a pack that lasts and one that fails is not expensive equipment or rare expertise. It is the discipline to match cells, prepare surfaces, test parameters on sacrificial material, and verify every connection. The energy required to form a reliable weld nugget occupies a narrow window, and finding that window is a methodical process, not a creative one. Respect the math, test before you commit, and build packs that outlast the devices they power.