Key takeaways
- Write the checklist and its numbers into the contract before the build starts: output, first-pass yield, unplanned stops, output interval, safety functions and documents. A limit agreed during the test is a negotiation, not an acceptance criterion.
- Measure output over a long, continuous window with stops included, for example 16 hours in eight two-hour blocks, and count each finished unit once at the line exit. A short demonstration at the fastest station proves little.
- Count first-pass yield with every quality cause and never put reworked units back into it. In the worked example below, a line that ended with 153 of 154 packs accepted still had a first-pass yield of 96.75% and failed.
- Test safety functions by challenge, not by looking: emergency stops, guard restart, power and air loss. One unexpected restart is a critical finding that puts the whole test on hold.
- A failed first run is normal. Log each nonconformity with severity, owner and deadline, retest what the fix touched, and repeat the full performance run when the fix changes line flow or counting.
- Technical acceptance, payment and shipment are separate decisions, and a FAT pass is neither a product certificate nor a site acceptance test.
Download the template
Download the FAT checklist (Excel). It has the 30 checks below with an evidence column, the acceptance limits, a blank performance-run sheet that calculates output, yield, stop ratio and OEE as you type, the run data from the worked example on this page, a nonconformity register and a sign-off record. Use and adapt it freely.
FAT, SAT and pre-shipment inspection
| Test | Where and when | What it shows | Typically tied to |
|---|---|---|---|
| Factory acceptance test (FAT) | The supplier’s factory, before the equipment is crated | The machine or line meets the agreed specification: safe, at rate, with traceable quality and complete documents | The payment that releases shipment |
| Site acceptance test (SAT) | Your plant, after installation and commissioning | It works with your utilities, materials and connected equipment | Final acceptance payment |
| Pre-shipment inspection | The factory or warehouse, when goods are packed | A sample of finished products matches the specification, quantity, marking and packing | Catalogue products, cable, components |
IEC 62381, revised in 2024, sets out requirements and checklists for the FAT, SAT and site integration test of process-industry automation systems, so that owners and suppliers agree on the scope of testing and who does what. It does not set output or yield numbers. Those come from your contract.
The checklist: 30 checks in three groups
Each check has a pass criterion and the evidence to attach. Enter N/A only with a written reason and the buyer’s approval. The process group is written for a battery module and PACK line; for a single machine such as a laser welder, keep the safety and handover groups and replace checks 11–18 with tests on your own parts at the settings you were quoted.
Configuration and safety
| Check | Passes when |
|---|---|
| 01 Contract baseline | Approved layout, bill of materials, product drawings, critical features and revisions match what is installed |
| 02 Identity and substitutions | Serial numbers and key component brands match the contract; every substitution has written approval |
| 03 Utilities and interfaces | Actual power, frequency, air and network conditions are recorded; upstream and downstream handshakes work |
| 04 Tools and metrology | Calibration dates, range and resolution of the torque, leak, electrical and weld test equipment suit the approved tests |
| 05 Guarding and access | Guards, closed laser enclosures and access control follow the reviewed safety design |
| 06 Emergency stop | Each stop circuit reaches its defined safe state; reset alone does not restart motion |
| 07 Guard restart sequence | Closing a guard, resetting and a deliberate start remain three separate actions |
| 08 Utility loss and recovery | After a controlled power or air loss the line recovers by the approved sequence, with no unintended restart |
| 09 Reject and buffer logic | A defective unit cannot advance as good; blocked and starved stations keep each unit’s identity |
| 10 Dry run and modes | Manual, setup and automatic modes, alarms and recipe selection behave as approved |
Process and traceability
| Check | Passes when |
|---|---|
| 11 Cell identity and polarity | Cell-to-module-to-pack genealogy is complete; wrong-polarity and wrong-identity challenges are caught |
| 12 Weld qualification | Joints pass the approved drawing, test method and force limit; each joint type has its own qualified limit |
| 13 Weld monitoring | Recipe, joint identity, alarms and reject status are recorded for every weld; strength is not judged by appearance |
| 14 Fastening records | Every specified fastener has a traceable torque and angle result before the unit is released as good |
| 15 Leak testing | Test pressure and limit are as approved; known-good and known-leak masters are classified correctly |
| 16 Electrical end-of-line test | Approved polarity, insulation, dielectric and functional tests, run by qualified people |
| 17 Duplicate serial control | A repeated identifier is rejected; an old pass record cannot be reused for a new unit |
| 18 Network and database loss | Records survive an outage; recovery reconciles units without duplicates |
| 19 Throughput and stability | The full run is witnessed with its real duration; unplanned stops and all quality causes stay in the report |
| 20 Recipe changeover | Last-good to first-good time is recorded and recipe interlocks work, against an agreed target |
Handover and shipment
| Check | Passes when |
|---|---|
| 21 English HMI and alarms | Every screen, alarm and reset instruction is in approved English, on the machine and in the manual |
| 22 Software handover | PLC and HMI project files, versions, licences, backups and access responsibilities are documented |
| 23 Backup restoration | The agreed restore method works and recipes and traceability data are valid afterwards |
| 24 Remote access | The access method and permissions are approved by the buyer; no undocumented default account |
| 25 Drawings and manuals | As-built electrical, pneumatic, mechanical and layout files match the final configuration |
| 26 Spares and training | Wear parts and special tools are counted against the list; operator and maintenance training scope is recorded |
| 27 Compliance evidence | Certificates and test reports for the destination market are checked for scope and issuer |
| 28 Packing and preservation | Split points, lifting points, transport fixtures and protection are approved |
| 29 Installation and SAT | Site utilities, interfaces, commissioning duties and the tests left for the SAT are written down |
| 30 Commercial release | The buyer alone authorises the payment and shipment milestone, after the required items are closed |
The numbers to agree before the build
A checklist without numbers turns the test into an argument. Agree these limits in the technical agreement, before manufacturing starts. The example values are for a storage PACK line designed for 8 cells per minute, as in the worked example below.
| Criterion | Example limit | How it is measured |
|---|---|---|
| Sustained gross output | At least 7.60 cell equivalents per minute, i.e. at least 152 completed 48-cell packs in 960 minutes (95% of the design rate) | Unique, first-time completed units at the final line exit over the whole window, stops included |
| First-pass yield | At least 99.00% | All quality causes count; a reworked unit never returns to the first-pass figure |
| Unplanned stop ratio | No more than 2.00% of the window | Stop minutes divided by window minutes |
| Output interval, 95th percentile | No more than 390 seconds (design: 360 seconds) | Nearest-rank 95th percentile of all intervals between completed units, stops included |
| Safety and traceability | Every safety challenge passes, no unexpected restart, every unit traceable | Challenge log and genealogy export |
| Closure | No open critical or major nonconformity | Nonconformity register |
Line speed in China is usually quoted in PPM, and the unit matters: at 8 cells per minute, a 48-cell pack should leave the line every 360 seconds. Agree whether PPM counts cells, modules or packs before you compare quotes. The percentile limit catches what an average hides: in run A below, the mean interval was 374 seconds, close to design, but the slowest 5% of intervals took 480 seconds or more because the buffers kept stalling.
Worked example: a storage PACK line that failed, then passed
This worked example shows what a complete FAT record and its calculations look like for a storage PACK line. It is not presented as a specific customer’s result.
The line makes packs of 48 LFP cells of 314 Ah: three 16-cell modules in series (1P48S), 153.6 V and 48.23 kWh nominal. Its scope runs from cell identification and sorting through module assembly, enclosed busbar laser welding, pack integration, fastening, leak test and end-of-line test, with traceability and reject handling throughout. Each performance run lasts 960 production minutes in eight two-hour blocks, with the same crew and a frozen recipe, hardware, software versions and tool calibration. Startup, safety challenges and destructive weld-coupon tests are done separately. For the kind of line this describes, see battery module and PACK lines.
| Measure | Run A, day 0 | Run B, day 3 |
|---|---|---|
| Completed packs | 154 | 158 |
| First-pass good | 149 | 157 |
| First-pass yield | 96.75% | 99.37% |
| Cell equivalents per minute | 7.70 | 7.90 |
| Unplanned stops | 30 minutes (3.13%) | 7 minutes (0.73%) |
| Output interval, 95th percentile | 480 seconds | 378.95 seconds |
| Decision | Hold | Performance limits met |
Run A failed on four counts. Gross output passed, but first-pass yield, the stop ratio and the 95th-percentile interval missed their limits, and an extended restart challenge found that closing a guard started the line in 1 of 12 attempts. Five packs failed first pass: two weld alarms, one missing fastening result, one leak-test reject and one duplicate serial number accepted at end of line. Four were reworked and one was scrapped, so 153 of 154 packs were eventually accepted, but the first-pass yield stays 149 of 154.
The nonconformities, their fixes and retests:
| Report | Severity | Finding | Correction | Retest |
|---|---|---|---|---|
| NCR01 | Critical | Guard closing restarted the line in 1 of 12 challenges | Separate reset and start logic; freeze the corrected safety configuration | 20 of 20 restart challenges and 40 of 40 safety regression checks passed |
| NCR02 | Major | Two weld alarms; 2 of 30 coupons below the 450 N reference limit | Correct clamp alignment, requalify the weld recipe | 30 of 30 new coupons passed, minimum 496 N |
| NCR03 | Major | A pack released without its fastening result | Release interlock requires every fastening record | 10 missing-record challenges blocked |
| NCR04 | Major | Leak-test reject from a worn fixture seal | Replace the seal, validate the master procedure | Good and leaking masters 20 of 20 each classified correctly |
| NCR05 | Major | A repeated pack ID accepted at end of line | Enforce unique IDs, block reuse of a pass record | 10 of 10 duplicate challenges blocked |
| NCR06 | Major | Stops 3.13% and 95th-percentile interval 480 seconds | Correct the buffer handshake and recovery sequence | Full run B passed |
| NCR07 | Minor | Two alarm reset instructions missing from the English manual | Revise the alarm appendix | Closed at the day 4 document review |
The 450 N weld limit is a reference value for one joint design; every joint needs its own qualified limit.
Run B passed. After corrections on days 1 and 2, the full 960-minute run was repeated on day 3. One pack failed first pass with a cosmetic defect on an incoming enclosure, and it stays in the yield figure. The English alarm appendix was closed at the document review on day 4, and technical acceptance followed. Packing inspection, shipment authorisation and the SAT remained open, as separate gates. The 157 of 158 result describes this 16-hour window; it is not a long-term yield guarantee.
How the run figures are calculated
Run A as an example, with an ideal cycle of 6 minutes per pack:
| Figure | Formula | Run A |
|---|---|---|
| Gross output | Completed packs ÷ window minutes | 154 ÷ 960 = 0.160 packs per minute |
| Cell equivalents | Completed packs × 48 ÷ window minutes | 7.70 per minute |
| Unplanned stop ratio | Stop minutes ÷ window minutes | 30 ÷ 960 = 3.13% |
| Availability | (Window − stops) ÷ window | 930 ÷ 960 = 96.88% |
| Performance | Completed packs × ideal cycle ÷ run time | 154 × 6 ÷ 930 = 99.35% |
| Quality (first-pass yield) | First-pass good ÷ completed | 149 ÷ 154 = 96.75% |
| OEE | Availability × performance × quality | 93.13% |
OEE is useful for finding where time goes, but separate limits on output, yield, stops and interval are clearer in a contract than a single OEE target. The template’s run sheet does these sums for you.
Witnessing a FAT by video
Many buyers cannot fly to China for every test. A video FAT works when it is planned as one:
- Before the day: get the protocol, the list of frozen versions and settings, and the camera plan. Agree which tests you watch live.
- Live: every safety challenge (emergency stops, guard restart, power and air loss) and the start and end of each performance block, with the line clock in view.
- Counts: a work-in-progress count at the start and the end of each run, so packs made before the window are not counted.
- Afterwards: raw exports rather than screenshots: serial genealogy, torque, weld, leak and end-of-line data, the downtime log and the time each unit left the line. With them you can recount the run yourself.
- The record: timestamped video segments referenced against each check, so someone who was not there can reconstruct the test.
When a FAT fails
Put the test on hold and open a nonconformity report for each failure: severity, finding, correction, owner, deadline and retest method. Decide what to retest by looking at what each fix changed: repeat the affected safety, quality and control checks, and repeat the full performance run when a change affects line flow or how units are counted. Record the changed configuration, the retest result and the reviewer’s approval for every closed report. Sign off only with no open critical or major item; minor items need an owner, a deadline and an agreed way to close them.
FAT, payment and shipment
- Keep three signatures apart. Technical acceptance says the equipment met the protocol. Releasing payment and authorising shipment are the buyer’s commercial decisions, made after it.
- Tie the pre-shipment payment to the signed record. For lines bought in China a common split is 30% advance, 30% on shipment after factory acceptance, 30% after final acceptance and 10% after warranty; export projects usually pay more before shipment. The storage PACK line cost guide has typical timelines, and our payment terms page covers letters of credit and other options.
- A FAT pass is not a certificate. CE documentation, Saudi Arabia’s SABER certificates and other market approvals are separate; check them under item 27.
- Shipment comes after more gates. A line is dismantled and crated after the FAT, the packing is inspected, and the SAT at your plant follows installation.
How we run FAT on our machines and lines
Every machine and line we supply runs a factory acceptance test against a checklist agreed in your contract, with your parts or cells where possible. You attend in Wuhan or watch live on video, and your own engineers or an inspection company you appoint are welcome. You receive the signed record, or a fix list and a retest; nothing is crated before sign-off, and the balance payment follows a passed test as agreed. Send us your line scope and we draft the protocol and the acceptance limits with the quotation. More on our factory acceptance and inspection service.