Assembly traceability and serialisation: lot, serial and test records

What traceability has to answer

Traceability is the ability to follow a part forward and backward. Backward means knowing which components, lots and process conditions went into a unit. Forward means knowing which units received a suspect lot and where they were shipped.

The commercial reason is scope. When a component lot is found to be out of specification, the difference between a recall of forty units and a recall of four thousand is the traceability record. The quality reason is root cause. Without the process data attached to the unit, an investigation becomes a review of averages.

The requirement should be written as questions the system must answer. Which coil lot went into serial 1042? Which torque was applied at station four? Which units were built on the afternoon shift with the batch of seals that was later rejected? A traceability design that cannot answer those questions is a filing system, not a traceability system.

Lot, batch or serial

Lot traceability records a group of units made from the same input lots in a defined window. It is cheap and simple, and it limits a recall to the group. The weakness is that a group may be large, and a single bad unit within it forces the whole lot to be treated as suspect.

Serial traceability gives every unit a unique identity. It allows a precise recall and a full history per unit, but it costs more: labels or marking, scanning, data storage and a process that keeps the serial attached through every step. It also creates the risk of orphan units, where a serial is never linked to a shipment.

Batch traceability sits between the two. It is useful when the process runs in defined batches, such as a potting or curing oven load, and when the unit itself is not serialised. The batch record can carry the oven curve and the material lot, and the units can be linked to the batch by count and time.

Designing the serial number

A serial number should be unique, stable and readable by a human and a machine. A common format uses a date or a Julian code, a line or cell code and a sequence number. A check digit catches typing errors at the station and in the field.

Avoid embedding too much meaning in the number. A format that encodes the variant, the shift and the customer becomes unreadable when the variant changes, and it can leak commercial information. Keep the number as an identifier and store the meaning in the database.

The carrier matters as much as the number. A laser mark survives most environments but can damage a thin wall or a coating. A printed label is flexible and cheap but can peel, fade or be removed. A moulded-in feature is permanent but hard to change and may not be readable after painting.

Choose the symbology for the scanner and the environment. A two dimensional code carries more data in a smaller space and tolerates some damage, but it needs a suitable reader and a clean print. A linear barcode is easier to read at a distance but carries less. Whichever you choose, define the quality grade and check it on the production part, not on a paper sample.

Capturing data at the station

The chain is built at the station. Each step should scan the serial, confirm the component or the operation and write a record. A scan that only opens a screen is not traceability; the record has to include the result.

Design the scan so it is easier than not scanning. Put the reader where the part already is, make the screen show the next action and block the cycle until the record is written. If the operator has to walk to a terminal, the scan will be skipped when the line is busy.

Capture the process data that matters: torque and angle, press force and distance, oven temperature and time, test voltage and result, leak rate. A pass or fail flag is useful, but the measured value is what allows an investigation. Store the units and the calibration status of the instrument.

Plan for the network to fail. A station should buffer records locally and upload when the connection returns, with a timestamp and a station identifier. A traceability system that stops the line every time the network drops will be bypassed.

The records that make up the chain

The table lists the record types and what each one should contain. The set is more important than the format, because the format can change with the system.

Labels that survive the journey

A label has to survive handling, cleaning, heat, cold, oil and time. A paper label with a water based adhesive will not survive a wash. A laser mark on a dark plastic may have poor contrast. A moulded-in code may be unreadable on a textured surface.

Test the label on the real part and the real process. Apply it, run it through the cleaning and the thermal cycle, then scan it. Check the grade with a verifier if the customer requires it, and record the grade on the label specification.

Think about the reading position. A code on the bottom of a unit that sits flat on a conveyor cannot be read. A code under a cable or behind a bracket is useless in service. Put the code where a service engineer can reach it with a handheld reader.

Have a fallback. If the label is damaged in the field, the unit should still be identifiable from a second mark, a moulded feature or the packaging record. A single label is a single point of failure.

Prove the chain with a mock recall

A traceability procedure that has never been run is a hypothesis. Pick a serial at random and ask the team to reconstruct its history: components, process data, test result, pack and shipment. Time the exercise and record where the chain breaks.

Then run it forward. Take a component lot and list every unit that contains it, then every carton and every customer. If the list cannot be produced in a useful time, the system is not ready for a real event.

Check reconciliation. The number of serials issued should match the number built, tested, packed and shipped, allowing for a known scrap and rework count. An unexplained gap is either an orphan unit or a record that was not written.

Review the result with the people who run the stations. The fix is often a scan position, a screen or a buffer, not a new database. Traceability is a line side behaviour before it is an IT project.

Retention, change control and access

Decide how long the records are kept and who can read them. The retention period should cover the product life, the warranty and any regulatory requirement. A record that is deleted after two years cannot support a ten year warranty claim.

Link the drawing revision to the build record. When a drawing changes, the cut-in serial or date should be recorded so an investigation can tell which units were built to which revision. A change that is not linked to the serial range is invisible in a recall.

Control access without making the data unusable. Quality, engineering and service need to read the records, and the customer may need a subset. Define the fields that can be shared, and keep the rest controlled.