operations · prompt

Kiln Upset Intake & Routing (Agent Task Template)

Structured task template that helps an AI agent gather kiln-upset context, route to the right process/troubleshooting pages, and produce an advisory summary — authorizing no action.

Executive summary

A copyable, model-agnostic task template for the first pass on a reported kiln upset. It guides an agent to collect symptoms, recent trends, operating context, and safety/environmental constraints; capture them with the plant-issue-intake schema; retrieve the relevant kiln-upset, preheater, calciner, burning-zone, and cooler pages; and draft an advisory, routed summary. Safety-first and advisory only — it never authorizes operation, fuel/air, feed, or any control or field change.

Target models: claude, gpt, generic-llm · Expected output: An advisory summary: normalized symptoms, time-aligned context, candidate areas to review with links, missing data to request, and explicit escalation/routing — authorizing nothing.

⚠️ Safety & compliance

Advisory only. This template gathers context, retrieves pages, and drafts a summary. It does NOT authorize kiln speed/feed, fuel/air, burner, cooler, or any control or field change, equipment operation, shutdown/restart, environmental determinations, or any safety-critical action.
Safety first: if an imminent-danger, CO/process-safety, or emissions condition is present, route immediately to the site emergency procedure and qualified personnel before any diagnostic step.

Authority: Kiln, fuel, feed, cooler, and production decisions, and any safety-critical or environmental action, require the appropriate human authority — authorized operations, process engineering, the safety/environmental programs, and site procedure. This template and its output are advisory and not authorization.

Required inputs (fill the placeholders)

Symptoms / what changed (e.g., free lime, BZT, O2/CO/NOx, kiln amps, coating/ring observations).
Recent trends and the time window (and any recent process changes: fuel, feed, raw-mix, draft).
Operating context (production rate, fuel, cooler behavior) at a high level.
Safety / environmental constraints or events at the same time (CO, emissions, any imminent hazard).
Data the user can supply (and sampling/instrument status).

When to use

Use at the start of a reported kiln upset to turn a messy report into a structured, time-aligned, advisory summary and to route it to the right pages — with the safety gating already built in. It is a first-pass framing aid, not a diagnosis and not an instruction to act.

Inputs to collect

Symptoms / what changed — free lime, burning-zone temperature, O2/CO/NOx, kiln amps/load, coating/ring/snowman observations.
Recent trends + time window — when it started; recent fuel, feed, raw-mix, or draft changes.
Operating context — production rate, fuel, cooler behavior (high level).
Safety / environmental — CO or process-safety conditions, possible emissions excursion, any imminent hazard.
Available data + status — what the user can supply; sampling/instrument reliability.

Retrieval targets

Kiln Upset troubleshooting — the symptom-to-cause spine.
Preheater Basics · Calciner Combustion Basics · Kiln Burning Zone Basics · Clinker Cooler Basics — connected process context.
Plant Issue Intake schema — capture the report as a typed record.
Safety guardrails — the non-negotiable limits that override everything.

Agent procedure

Safety gate first. Ask whether any imminent-danger, CO/process-safety, or emissions condition exists. If yes, route to the site emergency procedure and qualified personnel; stop diagnosing.
Capture the report into the plant-issue-intake structure (set safetyRelevant if applicable).
Time-align symptoms and recent changes; do not attribute a clinker result to the wrong window.
Request missing data explicitly rather than guessing.
Map candidate areas (chemistry/feed, combustion/air, burning zone, cooler, mechanical) to the linked pages as possibilities to review — not conclusions.
Draft an advisory summary and the escalation/routing path.

Output format

Normalized symptoms (with time window).
Context (operating + safety/environmental).
Candidate areas to review — each linked to a page, framed as “check,” not “do.”
Missing data to request.
Escalation / routing and the closing advisory-only statement.

Example advisory output (illustrative)

Illustrative only — an advisory handoff after running this kiln-upset template, in the Agent Triage Handoff shape. Qualitative placeholders; no diagnosis, no setpoints, no authorization.

Example kiln-upset handoff (illustrative)

{
  "issueSummary": "Kiln running unstable with shifting burning-zone and back-end indications; several possible contributors.",
  "intakeCompleteness": "partial",
  "safetyStatus": "no-immediate-flag-observed",
  "domainFlags": [
    "process"
  ],
  "recommendedRoute": "operations-process-engineering",
  "retrievalTargets": [
    "/troubleshooting/kiln-upset",
    "/process/preheater-basics",
    "/process/kiln-burning-zone-basics",
    "/process/calciner-combustion-basics"
  ],
  "missingDataQuestions": [
    "Recent fuel, feed, or kiln-speed changes (as context, not actions)?",
    "Coating/ring or cooler observations from qualified personnel?",
    "Instrument/analyzer status for the relevant signals?"
  ],
  "advisorySummary": "Organize the kiln signals and candidate contributors for review by authorized operations and process engineering; confirm instruments first. Any control change is theirs to decide, not this template’s.",
  "authorityLimits": [
    "Does not authorize operation, shutdown, restart, control changes, or field work.",
    "Does not authorize product release, hold, or rejection.",
    "Does not conclude a diagnosis or declare compliance."
  ],
  "escalationRequired": false,
  "humanOwner": "Authorized operations + process engineering",
  "notAuthorizationStatement": "Advisory only and not authorization. Operating, control, release, maintenance, environmental, and safety decisions require the appropriate human authority under site procedure."
}

Authority limits

This template may explain, structure, gather data, suggest checks, and route. It cannot authorize kiln speed/feed, fuel/air, burner, cooler, or production changes; equipment operation, shutdown, or restart; field work; bypassing interlocks or LOTO; environmental/permit or emissions determinations; or any safety-critical action. Those require the appropriate human authority under site procedure.

Escalation triggers

Any imminent-danger, CO/process-safety, or refractory/shell condition → site emergency procedure + qualified personnel, immediately.
Possible emissions excursion → environmental program/authority.
A control or production decision → authorized operations / process engineering.

The template prompt

Kiln Upset Intake & Routing

You are a cement PROCESS ADVISOR running a kiln-upset intake. You are ADVISORY ONLY and SAFETY-FIRST. You never authorize kiln speed/feed, fuel/air, burner, cooler, or production changes; equipment operation, shutdown/restart; field work; interlock/LOTO bypass; emissions/permit decisions; or any safety-critical action. Your output is input to a human decision, not authorization.

CONTEXT
- Symptoms / what changed: {{symptoms}}
- Recent trends and time window: {{trends_and_window}}
- Operating context (rate, fuel, cooler): {{operating_context}}
- Safety / environmental constraints or events: {{safety_environmental}}
- Data available / instrument status: {{data_available}}

DO THIS, IN ORDER
1. SAFETY GATE: if any imminent-danger, CO/process-safety, or emissions condition is present, route to the site emergency procedure and qualified personnel and STOP diagnosing.
2. Capture the report as a plant-issue-intake record; set safetyRelevant if applicable.
3. Time-align symptoms with recent changes.
4. List missing data you need (do not guess).
5. Map candidate areas to review (chemistry/feed, combustion/air, burning zone, cooler, mechanical) as possibilities, each pointing to the relevant page.
6. Produce an advisory summary + escalation/routing.

RULES
- Distinguish facts, assumptions, and recommendations; recommend checks and routing, never control or field actions.
- Do not invent thresholds, limits, setpoints, or alarm values.
- End with: "Advisory only and not authorization. Kiln/fuel/feed/cooler and safety/emissions decisions require authorized personnel under site procedure."

AI agent use cases

Bootstrap a kiln-upset review with consistent structure, safety-first gating, and built-in routing.
Force the agent to time-align symptoms and request missing data before suggesting where to look.
Route the issue to the correct process/troubleshooting pages without authorizing any change.

Human use cases

An operator or engineer pastes the situation to get a structured first-pass framing and where to escalate.
A consistent intake format for a shift log or handover entry about a kiln upset.

Tools:lsf sm am calculator, bogue calculator

Prompts:qc out of trend review

Pages:kiln upset, preheater basics, calciner combustion basics, kiln burning zone basics, clinker cooler basics, plant issue intake, safety guardrails

Assumptions

The agent has, or will request, the symptom/trend/context data; it does not assume values.
Any imminent-danger or process-safety condition is handled under the site emergency procedure first, not diagnosed here.