Automated fuel cell stack production lines offered by Chinese OEMs target first-pass equipment qualification rates ≥98%, with stacking CCD positioning accuracy of ±0.1mm, press stroke accuracy of ±0.02mm, and press accuracy grade of 0.2% FS [S1]. Lines cover both graphite-plate and metal-plate stacks, with air-cooled and liquid-cooled variants released as separate equipment families [S7][S8].
The practical scope of a "fuel cell stack production line" spans four station types: MEA/bipolar plate/end plate assembly, multi-stage stack compression, in-line leak detection, and screw/strap/tie-bar fastening. The output tier is set by the stack format chosen (core stack vs full system), the throughput target, and whether the line interfaces with an MES data backbone and AGV material flow [S2][S5].
Core Station Modules and Their Spec Floors
The seven station blocks that repeat across published line architectures are: feeding, stacking, pressing, fastening, in-line leak test, NG-station, and MES-linked outfeed [S1][S2][S4]. Imported high-precision robots paired with vacuum suction cups handle bipolar plate pick-up, with stacking tooling and CCD vision providing secondary fine positioning to ±0.1mm and positive/negative polarity recognition to prevent misplacement [S1].
Pressing stations support two operating regimes that specifiers should treat as mutually exclusive defaults: displacement-control (set stroke with a maximum pressure clamp) and pressure-control (set force with a maximum displacement clamp), with both limits interlocked and admin-locked [S4]. The published press accuracy numbers split into two camps: 0.2% FS with ±0.02mm stroke on Cube/Cubeenergy-class lines [S1], and ±1% FS on Mingzheng-class lines that pair this with a stated pressure-plate flatness of ≤0.1mm and a detection accuracy of ±(0.8% reading + 0.2% full scale) [S2].
Fastening is configurable as screw, tie-bar, or side-plate variants selectable per process recipe [S1]. Leak detection integrates either as a pressure-decay test or a flow-rate test, and can be selected per station as differential pressure or flow [S2][S4]. For a deeper comparison of upstream stack assembly equipment, see this Fuel Cell Stack Assembly Equipment spec map.
Stack Format Selection: Core Stack vs Full System
OEM catalogues split the market into three product families that map cleanly to different factory footprints: (a) core-stack assembly lines (Mingzheng, Shuntian, Toocle, Cubeenergy), (b) liquid-cooled stack assembly lines (Jinchen), and (c) fuel cell system intelligent assembly lines that include BOP integration (Jinchen) [S2][S3][S6][S7][S5].
The core-stack lines target the cell stack itself with stated output of 2 stacks/h, stacking precision of 0.1mm, press accuracy of ±1% FS, detection accuracy of ±(0.8% reading + 0.2% full scale), and upper/lower pressure-plate flatness of <0.1mm [S2]. Maximum stack size accommodates up to 400 cells, with an opening height of 1100mm for the press frame [S2].
Liquid-cooled stack lines add functional modules for MEA, bipolar plate, conductive plate, end plate, stack compression, stack leak detection, screw insertion, and nut locking, all on a single MES-linked control backbone [S7]. Air-cooled stack lines replicate that scope but with product handling sized to the smaller air-cooled stack form factor [S8]. A typical published footprint for a core-stack line is 29m length × 6m width, with manual assistance retained at select key stations to balance throughput and precision [S3].
Capacity, Footprint, and Throughput Specs Compared
Two published throughput benchmarks anchor the spec range. Mingzheng quotes 2 stacks/h for its core stack line, with throughput gated by the number of cells per stack [S2]. Jinchen quotes 1000 units/year for system-level assembly at stack power ≥150 kW, with a stated power range of 100-300 kW, product dimensions of 500×208mm and 305×150mm, and system transfer by AGV or human-assisted [S5].
The 2 stacks/h figure is a baseline for stacks with a moderate cell count and does not scale linearly with stack count; specifiers should treat it as a per-station cycle time, not a daily output. The 1000 units/year figure is a system-level, multi-shift planning number. The two numbers are not directly comparable because one is station cycle and the other is annual system throughput. For a fuller capacity-planning view across stack, BOP, and sourcing, see this stack capacity planning spec map.
Automation Tiers: Standalone, Semi-Auto, Fully Auto
Published line architectures present three discrete automation tiers that map to different capex envelopes. Standalone stack press machines support prototyping and low-volume runs, with manual/automatic modes selectable via a hardware switch, admin-locked parameter access, and interlocked pressure/displacement limits [S4]. Semi-automated lines retain manual loading at the press station while automating screw feed, tightening, and leak test [S4].
Fully automated lines integrate AGV-fed loading, vision-guided robotic or gantry stacking, automatic screw feeding and tightening, incline leak testing, and AGV outfeed, with all material movement tied to an MES backbone for full data traceability [S2][S4]. Cubeenergy-class lines add an NG (no-good) station dedicated to removing defective plates mid-stack without halting the line, keeping the upstream cycle continuous [S1].
Material flow in fully automated configurations uses AGV carts, with Cubeenergy supporting optional AGV distribution for lights-out operation, and Jinchen offering AGV or human-assisted transfer at the system level [S1][S5]. For a deeper look at how automated stack assembly fits inside a broader manufacturing execution flow, see this automatic molding line reference.
Traceability, MES Integration, and Quality Data
All six OEM datasheets surveyed put MES at the centre of the data architecture. The full process is data-traceable: production data is uploaded and bound in real time via a code-scanning system, with quality traceability tied to individual plates and stacks [S1][S2]. System-level lines add automatic model identification, recipe switching, and bidirectional task dispatch between MES and equipment [S5].
System-level test coverage on a Jinchen-class line includes system insulation testing, system communication detection, system weight and volume detection, and auxiliary automatic feeding [S5]. The line also includes electrical testing, performance testing, rework stations, safety warning light stacks, and a dedicated re-work path, all under a single electrical and data acquisition control layer [S5].
For in-line leak detection, the published options are pressure decay and flow rate, both integrated with flow meters, pressure sensors, controllers, regulators, valves, and a PLC-based control unit for real-time monitoring and traceability [S4]. Core-stack lines also offer a per-station differential pressure or flow detection selection, with air-tightness test performed after fastening, on the same press [S1][S2].
Decision Criteria: Who This Is For, and Where the Lines Break
A fuel cell stack production line is for an OEM or tier-1 system integrator that has already locked the stack architecture (cell count, plate material, cooling method) and needs a repeatable, traceable assembly process at ≥98% first-pass equipment qualification [S1][S2]. It is not for a research lab prototyping novel MEA chemistries, where a standalone stack press with admin-locked manual/auto modes is the more capital-appropriate choice [S4].
The published line breaks down into three risk vectors that specifiers should validate. First, the 0.1mm stacking accuracy is a positioning-accuracy figure, not a flatness or sealing-pressure uniformity figure; sealing quality still depends on plate flatness (≤0.1mm upper/lower) and the press accuracy grade [S1][S2]. Second, MES integration is consistent across vendors, but the fieldbus/protocol layer (how the MES talks to the PLC, robot controllers, and vision systems) is not specified in any of the surveyed datasheets and must be confirmed contractually. Third, the 1000 units/year figure assumes the stated stack power of ≥150 kW and a multi-shift operating model; lower power stacks or single-shift operation will pull the throughput down [S5].
Trackable signals to monitor over the next two quarters: (1) whether published press accuracy grades converge from the current 0.2% FS / ±1% FS split toward a common OEM benchmark; (2) whether more vendors publish the fieldbus protocol used for MES-to-PLC and MES-to-robot integration; (3) whether AGV-fed lines begin quoting lights-out run-time data alongside first-pass yield, given that AGV material flow is now standard across all surveyed fully-automated configurations [S1][S2][S5].
Spec-level background on the components involved: oxy fuel cutter, and load cell.