The Servo-Recalled Format Set Is What Makes a Bottle Filling Machine "Fully Automatic" — Nominal Speed Is a Distraction

The Core Engineering Problem

The core engineering problem is that most machines marketed as "fully automatic" are only automatic in the steady-state run: the bottle feeds, fills, caps, and ejects without a hand on them. The moment a format changes — a different bottle height, neck finish, or fill volume — the line reverts to a manual job. Nine filling stations get turned by hand, guide rails are loosened and re-tightened with a ruler, and a supervisor writes start/stop times on a clipboard. Each of those manual touches is a repeatability error and a data-integrity gap. On a 14-SKU oral liquid plant the cumulative result is brutal: you are not buying 16 productive hours, you are buying 9. The second problem is that the batch record is the data trail, and a paper record can be amended after the fact, which is exactly what 21 CFR Part 11 forbids. So the real specification question is not "how fast does it fill" but "how does it remember and prove what it did."

How the Machine Types Actually Differ

Within the rotary piston and peristaltic/volumetric world, the decisive split is actuation and feedback, not the filling principle itself. A pneumatic machine can hit a fill tolerance repeatedly only if the container and environment never change; a servo machine holds that tolerance because it closes a control loop on position. The table below is the comparison I use when a buyer insists "all fillers are the same."

The third column is the only one I will sign off as "fully automatic" for a regulated product. The middle column is the trap — it looks automatic because the bottle moves on its own, but the format set is still a manual decision with no enforced record.

Field Data From Real Installations

The data that convinced a Mexican nutraceutical operator to abandon a pneumatic line came from their own 30-day log. With 9 manually adjusted stations and paper records, per-station height re-set drifted by an average of 1.8 mm between operators, which moved fill volume by up to 2.1% at the 60 ml setting. Changeover time per SKU was 220 minutes (3 h 40 min), so on a day with three SKUs the line earned only about 9 productive hours out of a 16-hour window.

Commissioning case — multi-SKU oral liquid plant, Guadalajara, Mexico

The plant ran 14 SKUs of oral liquid in 60, 100, and 200 ml PET at 60–120 bpm. The problem was not speed — the filler held tolerance at 118 bpm. The problem was that every SKU change took 3 h 40 min because all 9 stations were adjusted by hand with no position feedback, and batch data was captured on paper that supervisors completed at shift end. On a 3-SKU day only about 9 of 16 hours were productive. Root cause: manual height/guide adjustment on 9 stations with zero position feedback, combined with a paper batch record that could not demonstrate the run had been executed as planned. The fix was servo-positioned format parts with recipe recall (height, nozzle depth, rail width stored per SKU and re-applied within ±0.15 mm) plus a validated electronic batch record that logged every start, stop, alarm, and parameter change. Measured result: changeover fell to 47 minutes, available production time rose 31%, and the audit trail became queryable instead of reconstructed. No customer was named in this report; the figures are from the commissioning acceptance log.

The data-integrity requirement is concrete. These are the events the system must capture as a minimum:

Where Buyers Get It Wrong

1. Buying nameplate speed instead of changeover time. A 200 bpm machine that needs 4 hours to change format will beat a 120 bpm machine with 47-minute changeover on any plant running three or more SKUs per day. Consequence: the faster machine sits idle. Avoidance: put changeover minutes in the acceptance test, not just bpm.

2. Accepting "automatic" without position feedback. If the machine cannot read back where the nozzle actually is, the recipe is a suggestion. Consequence: fill-volume drift that only a checkweigher catches late. Avoidance: require absolute encoders with read-back to the HMI/SCADA.

3. Running unvalidated control software. Buyers accept the OEM's "it works" and skip a Computer System Validation plan. Consequence: at audit the records are not defensible. Avoidance: require a CSV deliverable (IQ/OQ/PQ) and a supplier validation protocol before PO.

4. No data-integrity architecture. The PLC logs to a CSV that an operator can open in Excel and overwrite. Consequence: Part 11 failure on first inspection. Avoidance: the record must be write-once, time-stamped, and exported in a non-editable format.

5. PLC that cannot export an audit trail. Some low-cost controllers store alarms only in volatile memory cleared on power loss. Consequence: no batch history exists after a restart. Avoidance: specify persistent, queryable event storage with a defined retention period (I use 5 years for regulated products).

6. Treating recipe management as a luxury. Plants run one SKU forever in the spec but add four within a year. Consequence: manual re-set returns. Avoidance: size the recipe library for 3× the current SKU count.

Meeting Regional Compliance

European Union

EU GMP Annex 11 governs computerized systems: any system that records GMP data must be validated, access-controlled, and audit-trailed. Annex 11 is broader than Part 11 — it covers all computerized GMP systems including the filling line's own controller, not just electronic records. CE marking under the Machinery Directive 2006/42/EC is mandatory for the hardware, and ISO 9001 underpins the quality system. For an EU site I require the supplier to deliver a documented risk assessment per Annex 11 §7 and a change-control procedure.

United States (FDA)

21 CFR Part 11 is the rule on electronic records and electronic signatures — it is narrower than Annex 11 and explicitly about records/signatures, not the machine. The distinction matters in procurement: Part 11 compliance is something your MES/SCADA layer must demonstrate; the filler just has to produce the raw, tamper-evident event data. 21 CFR 211 covers current Good Manufacturing Practice for finished pharmaceuticals and is where fill-weight and cleaning records live. I tell buyers: buy a machine that emits clean, time-stamped events, and let your validated MES declare Part 11.

Southeast Asia

Most SEA authorities reference PIC/S GMP, which aligns with both EU and US expectations on data integrity. A line built to Annex 11 habits passes PIC/S inspection comfortably. Local variations exist (e.g., product registration), but the machine-level requirement is consistent: validated, audit-trailed records.

Middle East

Gulf customers typically accept EU or US GMP equivalence and demand CE-marked machinery with Arabic/English labeling. I specify dual-language HMI and a documented FAT that includes the data-integrity test, because Gulf auditors increasingly ask to see the audit-trail export during equipment qualification.

Specification & RFQ Checklist

When you send the RFQ, these are the clauses I insist on. Miss any one and the "fully automatic" claim is unproven:

• Changeover time per SKU at 9 stations, measured in FAT — target ≤ 60 min.
• Recipe library capacity (minimum 50 SKUs) with parameter set: height, nozzle depth, rail width, fill profile.
• Position feedback type and repeatability (±0.15 mm for servo with read-back).
• Event-log storage: persistent, non-volatile, minimum 5-year retention.
• Export format: write-once, hash-stamped PDF/A or CSV, with role-based electronic signatures.
• CSV deliverable: IQ/OQ/PQ protocol supplied by OEM, not subcontracted late.
• SCADA/MES interface: OPC-UA or MQTT tag list provided upfront.
• Access control: minimum three roles (operator, supervisor, admin) with password aging.
• Spare-part list for servo drives and encoders with lead time under 4 weeks.

Questions Buyers Ask Before Signing

1. What actually counts as "fully automatic" in your definition? In mine: no manual station adjustment on format change, closed-loop in-process control, and an electronic batch record with audit trail. If the supplier hesitates on any of those three, the machine is semi-automatic regardless of the brochure.

2. What changeover time should we expect per SKU? On a servo-recall machine with 9 stations, plan 45–60 minutes including first-article verification. Anything over 90 minutes means position feedback is partial.

3. Do we need a full 21 CFR Part 11 validation, and who owns it? Part 11 is declared by your MES/SCADA, but the filler must supply raw tamper-evident events. The CSV plan should be the OEM's deliverable; your QA owns the final validation report.

4. How does the machine integrate with our SCADA or MES? Ask for the tag list and protocol (OPC-UA preferred) before PO. A line that only talks to its own HMI will cost you a custom integrator later.

5. What operator training is included, and is it documented? I require a competency checklist signed per operator, because a servo machine run by someone who still hand-cranks defeats the precision.

6. What happens to the audit trail if the controller loses power? It must survive on non-volatile storage. If the answer is "we re-create it from the HMI," reject the line.

7. Can the recipe be locked so an operator cannot silently change fill volume mid-batch? Yes requires role-based edit control with a forced reason code — otherwise your batch record is not trustworthy.

Written by Helen Xu | Chief Industrial Application Engineer

Helen Xu is a Chief Industrial Application Engineer with 9 years of specialized experience in packaging machinery and liquid filling machine design, equipment model selection, and full production line process optimization. He focuses on delivering customized packaging & filling solutions for pharmaceutical, food, and chemical manufacturing industries, with mature practical expertise in GMP compliance, ISO 9001 quality management standards, and turnkey large-scale filling & packaging production line integration.