Drone component production: serial records and batch recalls

A customer reports that three quadcopters lost thrust in flight. The motors all came from your shop. Which winding lot did they use? Which magnet batch? Nobody can say without opening four spreadsheets and a stack of traveller sheets.

Most drone parts makers record the finished serial well enough. The trouble sits one level down. The stator was wound on a Tuesday shift, the bell was machined on another lot of aluminium bar, and the bearings came from two purchase orders. None of that was tied to the serial.

So the recall becomes a wide net. You pull every unit shipped in a quarter, because you cannot prove which ones are clean. Customers lose flying time. You lose trust.

Start with the genealogy for each serial. It lists every lot of every critical input that went into that unit: copper wire spool, magnet batch, bearing PO, carbon prepreg roll, PCB panel, MCU reel. One line per input, one parent serial.

Add the process records beside it. Cure oven profile for the frame, balance test result for the rotor, bench thrust curve for the motor, calibration file for the flight controller. These are the proof that a unit passed, not just that it shipped.

Keep supplier lot numbers as the supplier printed them. Do not retype them into a house format. When a vendor sends a notice about a bad reel, you need to search their number, not yours.

Kitting is the first weak point. A tray of parts goes to the bench, and the technician takes what is on top. If the lot is not scanned when the part leaves the store, the build record is a guess.

Rework is the second. A rotor comes back for rebalancing, a wire is resoldered, a board gets a new connector. The serial stays put. The history changes. If rework is not logged against that serial, the next failure looks inexplicable.

Mixed lots inside one bin are the third. Two reels of the same resistor sit in one drawer. Only one was bad. Without a rule to finish a reel before opening another, you cannot narrow the recall.

With a genealogy in place, the recall is a query. Enter the suspect magnet batch and get back every motor serial that used it. Then match those motors to the drone serials they shipped in, and to the customer and delivery date.

Test the logic before you need it. Pick one lot at random and trace forward to customers, then pick one customer unit and trace back to its raw inputs. If either takes more than an afternoon, the records have a gap.

Prepare the customer notice in advance too. Keep it plain. It needs the serial range, the symptom, the action asked of them and a return address. Writing it under pressure is how errors reach the field.

Do not try to trace everything at once. Choose the part that has failed most often or the one that sits on the most expensive airframes. Define five or six inputs that matter and record those first.

Run it for a month alongside your current method. Compare what the new genealogy tells you with what the old records tell you. Differences show where the habit needs fixing, usually at issue from stores.

Then add the next product. Ask your production head to pick one live order this week and write down, from memory, which lots went into it. The gaps in that list are your first job.