Quick start

Nothing here requires new hardware in the common case. What blocks pilots is access approval, so start those conversations first.

1. 1Deploy the edge gateway on the plant host and register it against your tenant.
2. 2Configure the first source: OPC UA endpoint, historian query, gauge folder or inspection database.
3. 3Confirm tag identity and timestamp behaviour against a known part or run.
4. 4Verify buffering by disconnecting the uplink and confirming records reconcile in order.

Registering a gateway

# Provided during onboarding: tenant URL and a single-use enrolment token
tqa-gateway enroll \
  --tenant https://acme.eu.trustqualityassurance.com \
  --token "$TQA_ENROLMENT_TOKEN" \
  --site "stuttgart-01" \
  --name "line-3-gateway"

# Verify connectivity and buffer state
tqa-gateway status --verbose

This is the week that determines whether later analysis is trustworthy. Characteristics are defined once, with the limits and sampling rule from your control plan rather than inferred from data.

1. 1Import or create the asset hierarchy: site, area, line, cell, station.
2. 2Define the products and operations in scope, with the drawing or recipe revision in force.
3. 3Define each characteristic: unit, specification limits, control limits, subgroup size and sampling rule.
4. 4Map source tags and inspection fields to characteristics; unmapped fields stay visible but unevaluated.
5. 5Have the process owner sign off the mapping — this is the artefact auditors will ask about later.

{
  "key": "seal-width",
  "name": "Seal width",
  "unit": "mm",
  "operation": "OP-40 sealing",
  "specification": { "nominal": 4.02, "usl": 4.30, "lsl": 3.75 },
  "control": { "chart": "xbar-r", "subgroupSize": 5, "ucl": 4.15, "lcl": 3.90 },
  "sampling": { "rule": "every-nth", "n": 20 },
  "ruleSet": "nelson-1-2-3-5",
  "criticality": "significant"
}

1. 1Enable control charts and capability tracking for mapped characteristics.
2. 2Configure alert severity, routing and escalation against your shift pattern and on-call ownership.
3. 3Enable vision or anomaly models where relevant, initially in shadow mode with no alerting.
4. 4Run operator onboarding: 30–45 minutes per shift, focused on what to do when an alert arrives.

1. 1Turn on NCR and CAPA workflows for the scoped area, with named owners and due dates.
2. 2Verify one full cycle end to end: alert, containment, root cause, action, effectiveness check.
3. 3Generate the evidence pack for the scope and confirm it satisfies your internal audit expectation.
4. 4Produce the pilot report: baseline versus measured result for each agreed success criterion.

Expansion is normally by line and then by site. The second line typically goes live in days rather than weeks because the characteristic library, connectors and alert routing patterns already exist. Multi-site governance — who owns shared definitions, how plant exceptions are handled — is covered in the rollout guide in the resource centre.