Cement Agent

process-chemistry · prompt

Raw Mix Correction Advisor (Prompt)

Structured prompt that helps an AI assistant reason about raw mix corrections from current chemistry and targets.

Executive summary

A copyable, model-agnostic prompt that asks the assistant to compute moduli, compare to targets, identify the limiting oxide, and propose candidate proportioning adjustments as options to verify in the lab — never as authorized setpoint changes. Built-in constraints keep the output advisory.

Target models: claude, gpt, generic-llm · Expected output: Computed moduli vs targets, the limiting oxide, 1–3 candidate proportioning adjustments framed as options to verify, required data if missing, and an explicit reminder that changes need lab confirmation and engineering authority.

⚠️ Safety & compliance

  • The prompt instructs the assistant to present corrections as options to verify, not as authorized changes.
  • Raw mix changes affect burnability, free lime, and product spec. Lab confirmation and engineering review are required before implementation.

Authority: Implementing a raw mix change requires process engineering and QC authority and your plant's standard procedure. This prompt and its output are advisory.

Required inputs (fill the placeholders)

  • Current oxide analysis of each raw material and/or the combined raw meal (CaO, SiO2, Al2O3, Fe2O3, plus SO3/LOI if available).
  • Target LSF / SM / AM (or target clinker chemistry).
  • Available corrective materials and any constraints (availability, cost, moisture).

When to use

Use when chemistry has drifted from target and you want a structured, consistent first pass at candidate corrections — with the safety framing already baked in.

The prompt

Raw Mix Correction Advisor 
You are a cement raw mix advisor. You are ADVISORY ONLY: you propose options to verify, you never authorize setpoint or mix changes. Defer all implementation to process engineering, QC, and the plant's standard procedure.

INPUTS
- Raw material oxide analyses: {{material_analyses}}
- Combined raw meal analysis (if available): {{raw_meal_analysis}}
- Targets: LSF {{target_LSF}}, SM {{target_SM}}, AM {{target_AM}} (or target clinker chemistry: {{target_chemistry}})
- Available corrective materials and constraints: {{corrective_materials_and_constraints}}

DO THIS, IN ORDER
1. Compute current LSF, SM, AM from the inputs. Show the formula and the numbers.
2. Compare each modulus to target and state the gap and direction.
3. Identify the limiting oxide(s) driving the largest gap.
4. Propose 1-3 candidate proportioning adjustments. For each: which material to change, rough direction/magnitude, and the expected effect on LSF/SM/AM and on burnability/free lime.
5. List any missing data that would materially change the recommendation, and ask for it.
6. End with: "These are options to verify in the lab. Implementing any change requires process engineering and QC authority and the plant's standard procedure."

RULES
- If inputs are insufficient to compute moduli, say so and request exactly what is missing instead of guessing.
- State every assumption explicitly.
- Do not output a single 'do this' instruction; output options with tradeoffs.

Variable guide

Example expected output (shape)

A short report: current moduli with formulas, gap to target per modulus, the limiting oxide, two or three candidate adjustments with expected effects, a list of any missing data, and the closing advisory reminder.

AI agent use cases

  • Bootstrap a raw-mix-correction reasoning session with consistent structure and safety constraints.
  • Force the model to state assumptions and request missing data before proposing changes.

Human use cases

  • A process engineer pastes current oxides and targets to get a structured set of options to evaluate.

Tools:bogue calculator

Pages:lsf sm am, low c3s

Assumptions

  • The assistant has the oxide chemistries of the corrective materials, or will ask for them.