What Should You Know Before Visiting a Battery Factory in China

May, June, September. The lunar new year disruption radiates outward from the official holiday dates by weeks in both directions. By mid-January skilled operators are already leaving for home provinces. March is retraining season. October Golden Week is a smaller version of the same problem.

Subsidy timing is harder to read from outside. When EV registration deadlines compress OEM schedules, the pressure hits cell factories as demands to ship faster, which means compressing every time-dependent process step. The tell is atmospheric: formation racks that look overfull, forklifts moving with unusual urgency, cells spending less time sitting between stations. Recognizing these signs requires a prior visit for comparison. Without a baseline, a first-time visitor cannot distinguish a factory under delivery pressure from a factory that always runs this way.

This section is long because the dry room is where the money gets lost. Not theoretically. Financially. Moisture contamination during cell assembly produces cells that are indistinguishable from good cells at the point of manufacture and then fail in the field a year later, inside assembled packs that cost a fortune to open.

This is why dry room audit technique matters more than any other observation during the visit, and it is also why most visitors get the dry room audit wrong. The typical approach is to walk in, glance at the dew point display on the wall, note that it reads negative 45°C, and move on. That reading is scenery.

Request the trend log. Four days of continuous data. The overnight hours on non-visit days are where the real performance lives. An excursion to negative 25°C at 3 AM means the dehumidification system faulted. Faults happen, even in well-maintained systems. What happened next is the question. Were the cells in assembly during the excursion flagged for quarantine? Were they scrapped? Or were they quietly returned to the production stream because nobody wanted to explain to management why the night shift lost four hours of output?

Molecular sieve rotor replacement cycles are worth asking about, and the reason almost nobody does is that the question sits in an uncomfortable interdisciplinary gap. Most visitors are either electrochemistry people who do not think about HVAC components, or supply chain people who do not think about either. The rotors are desiccant wheels. They adsorb moisture on one face while thermally regenerating on the other, cycling continuously. The desiccant media degrades over years. An eight-year-old rotor still functions but with reduced adsorption capacity and slower recovery from system upsets. When something else goes wrong in the dehumidification loop, an aged rotor is the difference between a ten-minute recovery and an hour-long one. Ask the facilities engineer when the rotors were last replaced. If the answer comes immediately with a date, the maintenance program is tracked at the component level. If the answer involves looking something up or asking someone else, it probably is not.

Now, standing near the dry room entrance. Watch operators transit the airlocks for ten minutes without speaking. Watch whether corrugated cardboard enters the room. Cardboard at ambient conditions holds 7-9% moisture by weight. Inside a dry room it releases that moisture into the local environment over hours. The room-average dew point may barely budge because the dehumidification system compensates globally, but the electrode trays sitting within a few meters of the cardboard are experiencing local humidity that the wall sensor does not measure. Factories that have truly internalized dry room discipline use polymer totes and metal carts exclusively and enforce it at the shift supervisor level. The absence of cardboard inside a dry room is not a detail. It is a cultural signal about how the factory thinks about contamination.

One physical check: hold a tissue near the door seal. It should push away from the dry room (positive pressure). If it pulls inward, ambient air is leaking in continuously.

The useful information at the coating station can be gathered in fifteen minutes, and spending much longer than that is a misallocation of limited visit time. The equipment looks impressive, the factory is comfortable being observed here, and neither of those facts correlates with audit value.

Cross-web coating weight CV, with SPC documentation. Below ±1.5% with a chart showing sustained statistical control is what to look for. Undocumented numbers are engineering targets, not measurements.

Scrap reclamation practice. If the factory reclaims scrapped electrode material by redissolving it into fresh slurry, metallic particle screening on the reclaimed batch needs to be demonstrated. Without screening, reclaim is a contamination recycling loop.

Line speed as a fraction of rated speed. The question goes to whoever is operating the equipment, not the tour escort.

The injection equipment is the less interesting half of this process step. Gravimetric dosing versus volumetric dosing is a meaningful distinction (gravimetric compensates for electrolyte density variation, volumetric does not), and it can be determined in thirty seconds by looking at the injection hardware.

Post-injection soaking is where the exposure actually lives, and it is also where most technical writing on this subject, including plenty of factory-issued process descriptions, understates the difficulty. Electrolyte needs to wet the full pore structure of electrodes and separator by capillary action. For large-format prismatic cells with thick, densely calendered cathodes, the wetting process can take well beyond 20 hours. Some cathode-separator combinations are genuinely difficult to wet, particularly with high-nickel cathodes calendered to porosities at the low end of the 25-35% range, where pore tortuosity slows capillary penetration substantially.

Count the formation channels. Ask the daily output claim and the formation protocol duration. Multiply. If the factory claims 150,000 cells per day and a 60-hour protocol, there need to be 375,000 cells' worth of formation capacity in that room. If there are 200,000 channels, the math does not work.

Compressed formation produces a lower-quality SEI. There is broad consensus in the electrochemistry literature on this point, though the exact relationship between formation protocol parameters and long-term SEI stability is still an active area of research and varies with anode chemistry, electrolyte formulation, and additive package. What is not in dispute is that faster initial charge rates produce thicker, less passivating interphase films that degrade cell performance over the full service life. The practical difficulty is that the degradation emerges gradually and is convolved with temperature effects, cycling patterns, and calendar aging, making it nearly impossible to trace a field performance shortfall back to formation protocol compression with certainty. This ambiguity is precisely what allows factories to get away with shortening formation.

Cells sit on racks for one to three weeks post-formation. OCV is monitored for abnormal self-discharge indicating metallic particle micro-shorts. Fourteen days is the automotive standard. Seven is a bare minimum for any serious application. Verify the stated duration against the dates visible on rack labels.

Grading sorts finished cells into performance bins by capacity, impedance, and self-discharge rate. The tightness of the grading bins determines how well-matched cells will be within a finished battery pack, which affects pack-level performance and longevity. Ask what the bin width is for capacity and impedance. Tighter is better, but tighter also means more cells falling outside the primary bin and being sold at discount into lower-specification markets.

Calibration stickers. Current or expired. Expired means drift is unmeasured on every result since expiration.

Failure analysis is a separate and larger question. Some factories have purchased expensive analytical equipment for the lab tour and generate almost no investigative reports. The relevant question is whether the factory performs root cause analysis on rejected cells and field returns, and whether those investigations produce documented process changes. Ask for a corrective action report from the past six months. A factory that can produce one has a feedback loop between failure and improvement. A factory that cannot has a disposal procedure instead of an investigation procedure, and the failure modes persist.

This station gets skipped on most tours and it should not. Cathode powder moisture, particle size distribution. Separator thickness, porosity, Gurley number. Electrolyte water content, HF level. These are the incoming material parameters that most directly affect finished cell quality. A factory that performs its own verification has a firewall against supply chain variability. A factory that relies on supplier certificates of analysis is extending trust to organizations it does not audit, in a supply chain where variability between batches and between suppliers is considerable.

Write out the full process flow before arriving. Mixing, coating, drying, calendering, slitting, electrode assembly, tab welding, casing, electrolyte injection, soaking, sealing, formation, aging, testing, grading, packing. During the tour, track which stations are shown. The omissions are where the problems tend to be. The three most commonly skipped stations across the industry are incoming material inspection, electrolyte injection, and the scrap/rework area.

The dinner surfaces intelligence that the tour suppresses. Technical staff, off-script and relaxed, will mention things with major supply continuity implications. Cathode supplier transitions. New lines still in qualification. Recent OEM audit findings that forced process changes. These details emerge because the social context lowers the commercial caution that governs daytime interactions, and because technical people often want to talk about the difficult problems they have solved.

There is a subtlety to dinner intelligence gathering that goes beyond simply "bring a technical person." The sequence of the conversation matters. Early in the dinner, when the atmosphere is still somewhat formal, questions about general capability and capacity are appropriate. The more specific and revealing disclosures tend to come later, after the formality has dissolved, and often arise not from direct questions but from tangential storytelling about recent challenges. A factory's chief engineer describing a difficult month with a new cathode lot is volunteering information about supply chain vulnerability that would never appear in an RFQ response. That information has direct bearing on whether the buyer's upcoming order will encounter the same problem.

IATF 16949 forces quantitative process capability measurement, FMEA, layered process audits. ISO 9001 requires a documented QMS without mandating quantitative capability analysis. For automotive applications, IATF 16949 is non-negotiable. For stationary storage, the certification itself matters less than the process discipline it indicates.

Pull from the line. Label independently. Record batch, line, shift, date. Ship to a third-party lab. Define test conditions in a written, countersigned protocol before leaving.

A second visit six months later is worth more than the first visit was, because the comparison reveals trajectory. Were corrective actions from the first visit sustained or forgotten? Is the workforce stable? Is equipment condition tracking with utilization or falling behind? The factory's ability to manage the visitor experience degrades across repeated visits, because maintaining a performance indefinitely requires the performance to become the actual standard. Semi-announced visits, where the factory knows the approximate window but not the specific day, are logistically practical and commercially accepted by confident suppliers.