Machine Data Acquisition in Regulated Manufacturing

Machine Data Acquisition is the controlled capture of equipment-generated measurements, states, and events, together with their source and process context, for use in manufacturing systems and records. It connects what equipment reports with information that manufacturing personnel can interpret. This is an operational definition, not a statutory one.

These functions can coexist within a system; they are not mutually exclusive product categories. A platform may collect signals, retain history, and coordinate execution. Naming the function makes its responsibility clearer than relying on the product label.

Equipment-originated observations differ from manual entries and derived calculations. A temperature transmitted by a controller is an equipment observation; a temperature typed by an operator is a manual entry, even if read from that controller. An average calculated from captured temperatures is a derived result, not another direct observation.

A captured value does not by itself establish its batch association, suitability for a quality decision, or approval status. Production context tagging addresses the association question; interpretation also depends on provenance and intended use.

Start with the product, jurisdiction, equipment function, and intended record—not the machine connection. This mapping concerns US drug products under 21 CFR 211. For supplements, food, devices, chemicals, and non-US operations, check the applicable product and jurisdiction rules before adopting it.

For US drug-product equipment under 21 CFR 211.68(a), equipment used in the covered operations must be routinely calibrated, inspected, or checked according to a written program designed to assure proper performance. For that US drug-product equipment, the same provision requires written records of those calibration checks and inspections to be maintained.

For US drug products, 21 CFR 211.188 requires batch production and control records for each batch, including complete information relating to its production and control. Operationally, define how acquired values contribute to the electronic batch record; collection alone is not a record-design decision.

For US drug-product electronic records in closed systems where Part 11 applies, 21 CFR 11.10 requires procedures and controls designed to ensure authenticity, integrity, and, when appropriate, confidentiality. Do not infer Part 11 applicability merely from a machine connection. See Electronic-record integrity and Part 11 controls for detailed controls, and Validation, change control, and lifecycle testing for acquisition testing.

Keep the categories separate: the cited provisions are binding law within their scope; official guidance recommendations remain guidance; interface and sampling choices are engineering practices; software support implements selected functions. Do not treat an engineering convention as a legal mandate or a software feature as a compliance determination. Human review and approval remain assigned responsibilities.

A practical acquisition architecture separates logical responsibilities, not necessarily servers or products. Components may be combined or deployed separately; the following workflow is an industry-practice model, not a prescribed standard.

Instrument or equipment controller (source identification) → acquisition adapter (event capture) → local event buffer (temporary retention) → transport service (transfer) → contextualization service (equipment and batch association) → historian or record repository (storage) → review interface (exception handling and human review).

Locate the acquisition boundary at the actual source interface: for example, the controller tag read by the adapter. That value may already be filtered, scaled, or calculated. Reading it does not necessarily expose the original sensor reading, calibration history, or intermediate calculations. Trace those dependencies when they affect the intended record; distinguish an acquired value from an independently measured value.

At each handoff, assign responsibility for both successful delivery and exceptions. Automation commonly owns source meaning and adapter configuration; IT commonly operates buffering and transport; manufacturing identifies execution context; quality defines review expectations. These are illustrative assignments, not a universal organization chart. Agree who resolves an unidentified batch, rejected event, or unavailable destination before commissioning. Use production context tagging to define associations and MES historian integration to clarify which repository serves each record purpose.

Read-only acquisition is a design option for observing equipment without sending operational changes. Command or recipe writeback is a separate interface risk to assess, with its own authorization and failure behavior. Connecting a reporting service does not, by itself, establish equipment-control authority. Draw observation and command paths separately so ownership remains visible.

OPC UA, MQTT, equipment-specific APIs, and file exchange are industry integration options—not regulatory controls or assurances of a particular product integration. For each implementation, evaluate actual field semantics, authentication, quality indicators, acknowledgments, and reconnect behavior. Establish whether an acknowledgment means transport receipt or application acceptance, and whether reconnecting returns current values, queued observations, or both. These distinctions shape MES–historian integration.

Choose capture behavior from the process question

• Polling reads at scheduled opportunities; it can repeat unchanged values and miss transients between reads.
• Change-driven capture reports qualifying changes; source-side filtering or deadbands can suppress transitions relevant to the intended analysis.
• Equipment-event capture records defined occurrences, such as completion of a weighing step; confirm exactly which transition generates the event.

Select the approach from the question being answered and the expected state transitions, rather than applying a universal sampling interval. Check whether the source exposes the necessary transitions before choosing downstream capture settings.

Preserve meaning at the interface

A practical interface contract identifies the channel, raw value, source unit, normalized unit, quality code, source time, receipt time, and mapping version. Source time describes the observation; receipt time describes its arrival. Define missing, stale, and invalid states separately from a valid zero. Keep any display or processing replacement identifiable as a substitute, not a measurement.

Worked mass calculation

Assume valid gross mass G kg and tare mass T kg refer to the same container and weighing event. Net mass is N = G − T, expressed in kg. Preserve both source-event references and the calculation version alongside N. An invalid or mismatched tare leaves the derived result unresolved; substituting zero would change the calculation's meaning. See tare-verified weighing for the weighing-specific context.

A useful acquisition record identifies not just a value, but what produced it and which execution it describes. Common practice uses stable event identifiers and explicit relationships for production context tagging, rather than embedding a batch name in an unstructured signal label. Equipment-channel identity distinguishes the physical asset and measurement point; effective process context identifies the batch, operation, and interval to which an event belongs.

For US drug-product batch records, 21 CFR 211.188(b)(2) requires the identity of individual major equipment and lines used. That record requirement is distinct from the engineering method chosen to associate machine events with equipment.

Logical evidence model

The relationship diagram below is illustrative practice, not a mandated schema. Arrows denote explicit links.

Interfaces, sampling, and unit normalization defines source time and receipt time; the same distinction applies here so transport timing is not treated as process timing. Versioned transformations allow a reviewer to reconstruct an interpreted value without applying today's mapping to yesterday's event.

Worked batch handover

Assume consecutive batches share equipment and flow is acquired in kg/h. A delayed event arrives during the later batch, but its source time is uncertain at handover.

• Retained event: preserve its identifier, flow value, channel, timestamps, and mapping version.
• Candidate context: record both adjacent batches as candidates; leave the association unresolved.
• Corroborating evidence: compare equipment execution transitions with batch execution history. Receipt during the later batch is insufficient by itself.
• Final disposition: if execution evidence places the event in the earlier batch, the reviewer records that association and rationale. Otherwise, retain unresolved status rather than force attribution.

Binding closed-system controls

For US drug-product electronic records in closed systems subject to Part 11, 21 CFR 11.10(e) requires secure, computer-generated, time-stamped audit trails independently recording the date and time of operator entries and actions that create, modify, or delete electronic records, and requires that changes not obscure previously recorded information.

For US drug-product electronic records in closed systems subject to Part 11, 21 CFR 11.10(b) requires the ability to generate accurate and complete record copies in human-readable and electronic form suitable for agency inspection, review, and copying.

For US drug-product electronic records in closed systems subject to Part 11, 21 CFR 11.10(c) requires protection enabling accurate and ready retrieval throughout the records retention period.

For US drug-product electronic records in closed systems subject to Part 11, 21 CFR 11.10(h) requires device checks, as appropriate, to determine the validity of the source of data input or operational instruction.

Practice: distinguish the evidence types

• Measurement history typically shows observed values, timestamps, units, and quality flags. It describes the process, not necessarily the operator’s changes to records.
• Transport diagnostics typically show connection failures, retries, and delivery status. They help explain missing or delayed measurements but do not establish who corrected a record.
• An operator-change audit trail records actions affecting electronic records. A continuously sampled signal history is not a substitute for that evidence.

Practice: export an interpretable record

A readable acquisition export can pair measurements with equipment and batch context, applicable tag-mapping references, and transformation versions. Correction lineage can connect the original value, corrected value, responsible person, time, and reason rather than presenting only the latest value. Linked review status and unresolved exceptions help a reviewer understand what was assessed.

A trend screenshot may support interpretation, but it is not inherently a complete record copy: it may omit underlying values, flags, corrections, or context. Define export completeness against the record’s intended meaning, then check retrieval of those relationships as part of record retention and archiving. As in Regulatory scope and expectations, generating an export supports review; it does not substitute for the human approval decision.

Binding access and backup obligations

For US drug-product electronic records in closed systems subject to Part 11, 21 CFR 11.10(d) requires system access to be limited to authorized individuals. For US drug-product computer or related systems, 21 CFR 211.68(b) requires controls restricting changes in master production and control records or other records to authorized personnel. For backup arrangements in US drug-product computer or related systems, 21 CFR 211.68(b) requires safeguards addressing backup data exactness, completeness, alteration, inadvertent erasure, and loss.

Engineering choices for continuity

In MES and SCADA integration, common engineering choices include network segregation to limit reachable systems and restricted service identities to limit connector permissions. These are implementation patterns, not mandated architectures in the provisions above.

• Durable buffering retains events during interrupted transport. Acknowledgment tracking distinguishes pending events from those the destination has accepted; define whether acceptance means receipt or durable storage.
• Idempotent replay prevents retransmitted events from creating duplicate destination records. Evaluate event identity across reconnects and equipment restarts.
• Recovery monitoring exposes backlog, failed deliveries, and unresolved gaps. Treat buffer exhaustion and source restart as explicit exceptions: exhaustion may discard events, while restart may reset sequence identifiers or erase untransmitted data.

A delivery buffer is not a backup. Replay addresses interrupted transport; restoring retained records after corruption or loss addresses a different failure boundary. For data integrity, evaluate transport recovery and record restoration separately, with named owners for investigating exceptions rather than treating a reconnected interface as proof of complete recovery.

In industry practice, acquisition-health metrics are meaningful only when the capture plan defines expected behavior, denominators, and the measurement window:

• Completeness: distinct expected events received divided by the expected event set for the window, accounting for scheduled sampling or event-triggered capture.
• Freshness: elapsed time from the source event timestamp to receipt, using a defined clock basis and identifying delayed arrivals.
• Contextualization coverage: eligible events with the required equipment, batch, or execution associations divided by all eligible events.
• Exception backlog: unresolved acquisition or context issues at the review cutoff, grouped by age, significance, and owner.

A growing row count alone does not establish completeness: duplicates can conceal missing events. Thresholds are typically chosen for the intended capture behavior and operational use, not a universal target.

Exception-based review commonly prioritizes human attention using documented selection rules, visible excluded data, and access to underlying events and transformations. An acquisition-health alert identifies a data-delivery or context issue, not a process-quality conclusion. Statistical process control (SPC) is a possible downstream use of suitable, contextualized measurements—not a substitute for assessing acquisition health.

For US drug-product operations addressed by 21 CFR 211.68(c), the automated-equipment checking arrangement is conditional on equipment conformity with that section and a person checking that the equipment properly performed the operation.

For US drug-product batch records involving a significant step performed by automated equipment under 21 CFR 211.68, 21 CFR 211.188(b)(11) requires identification of the person checking that step. This identifies a human checking responsibility relevant to electronic batch records, not merely a dashboard status.

For US drug-product electronic records in closed systems subject to Part 11, 21 CFR 11.10(a) requires system validation addressing accuracy, reliability, consistent intended performance, and the ability to discern invalid or altered records.

For US drug-product computer or related systems, 21 CFR 211.68(b) requires formulas, records, or data entering and leaving the system to be checked for accuracy; the degree and frequency of verification must be based on system complexity and reliability.

In computer system validation, an acquisition challenge matrix commonly connects expected results to intended use—not an arbitrary test count. These illustrative operational qualification challenges translate that approach into observable evidence:

Lifecycle assessment commonly treats connector configuration, equipment firmware, tag definitions, transformations, and context rules as change-impact inputs. Preserve the relationship between historical records and the configuration that interpreted them. Select regression challenges around affected behavior, including whether older records remain interpretable after a change.

This industry-practice example assumes temperature events in °C have stable source-event identifiers, source timestamps, and durable local buffering. The network fails after an event is accepted downstream but before its acknowledgment reaches the sender. On reconnection, replay includes that event.

Match the replayed identifier to the existing record and compare its payload before treating it as a duplicate. Investigate conflicting payloads rather than overwriting them. Reconcile the remaining buffered events against downstream records, retaining source timestamps and receipt timing in the process event log. Document any unrecoverable gap and its batch relevance. Buffering does not guarantee lossless operation: buffer exhaustion, corruption, or incomplete capture can still leave missing observations.

These symptoms guide investigation; they do not establish causes. Use production context tagging evidence to resolve batch association, not arrival order alone.

Reprocessing a retained raw event changes its interpretation; obtaining a new physical measurement observes conditions at a later time. Retain the original event and the basis for any changed interpretation, using the integrity controls covered above. Do not silently fabricate replacement observations or present a later measurement as evidence of the missing interval.

Use these practical readiness questions alongside the earlier architecture, integrity, recovery, and lifecycle discussions; they are not a substitute for determining applicable regulatory obligations.

• Is the intended record use defined, including which decisions depend on acquired data?
• Are source and channel identities controlled, with configuration ownership identified?
• Are units and transformation versions traceable from recorded results to source values?
• Are expected capture behavior and production context rules explicit, including treatment of missing or ambiguous context?
• Have recovery and export behavior been challenged against the intended record use?
• Can reviewers locate unresolved exceptions and supporting evidence through the exception-based review workflow?

Ownership and operational handover

Assign named roles for source configuration, interface changes, record interpretation, exception disposition, and operational support. Document where responsibility passes between equipment, integration, manufacturing, and quality teams. Use the site's approval structure rather than assuming a universal hierarchy.

Before routine operation, confirm that procedures are accessible, assigned personnel are trained, escalation routes are understood, and acceptance evidence is retrievable. Record outstanding limitations and who owns their resolution.

As established in Regulatory scope and expectations, software supports these checks but does not itself satisfy them; keep review, approval, and release decisions with authorized humans accountable for each.