Units
1. Introduction to Artificial Intelligence in Metallurgy: Boundaries, Validation, Responsibility and Ethics 2. AI in Material Selection and Decision Support Systems 3. Feature Prediction and Structure-Process-Property Relationships (Machine Learning) 4. Phase Diagrams, Thermodynamics and AI in CALPHAD Approach 5. Microstructure Image Analysis and Quantitative Metallography 6. AI in Optimization of Heat Treatment and Process Parameters 7. Manufacturing Processes: AI in Casting, Welding and Additive Manufacturing 8. Quality Control, Defect Detection and Statistical Process Control 9. AI in Damage Analysis, Root Cause and Breakage Investigation 10. Characterization Data Interpretation: XRD, SEM/EDS, DSC and Mechanical Testing 11. Standard Investigation, Compliance and Verification Discipline 12. Metallurgical Data Analysis and Frontiers of AI with Python
Unit 6 / 12

AI in Optimization of Heat Treatment and Process Parameters

Gains:

  • Ability to configure austenitizing, quenching and tempering parameters with AI according to the target feature and design an experiment matrix
  • Ability to analyze the relationship between hardenability (Jominy), CCT/TTT curves and tempering parameters with AI support
  • Ability to test AI's recommended heat treatment recipe with standard, furnace reality and validation testing

Two pieces made of the same steel can be one flexible like a spring and the other brittle like glass just because they are heat treated differently. Heat treatment is the process of changing the microstructure and therefore the mechanical properties of the material by heating and cooling it in a controlled manner. The typical cycle in steels consists of austenitizing (bringing to the austenite phase at high temperature), quenching (rapid cooling and forming martensite) and tempering (softening the martensite and gaining toughness) steps. The temperature and duration of these steps determine the targeted hardness and toughness. AI accelerates establishing an initial recipe based on the target feature, designing the experiment matrix, and interpreting the curves. However, furnace reality, section thickness and hardenability make the recipe different from paper; It cannot go into mass production until every recipe is verified in a trial piece.

Basic concepts of heat treatment and the role of AI

  • Hardenability: How deeply a steel can harden when quenched. The amount of carbon determines the level of hardness, and the alloying elements (Cr, Mo, Ni) determine the depth of hardening. The Jominy test quantifies hardenability by spraying water at one end of a standard sample and measuring how hardness decreases with distance.
  • TTT and CCT curves: TTT (Time-Temperature-Transformation: transformation at constant temperature) and CCT (Continuous Cooling Transformation: transformation in continuous cooling) curves show which microstructure (perlite, bainite, martensite) will form at what cooling rate. It is the map of recipe design.
  • Tempering balance: High tempering temperature reduces hardness but increases toughness. The goal is to achieve this balance according to the application.

AI is powerful at teaching these concepts, establishing what steps will be required based on a target hardness, and recommending an experiment matrix (different temperature/time combinations). But the actual position of the CCT curve depends on the exact composition of the steel; The AI's "memorized" curve is not your party's actual curve.

Tip: When evaluating a heat treatment recipe, always ask about section thickness. The quenching that works in a thin sample cannot reach the center in a thick section (mass effect); the center remains soft. Give the AI ​​the thickest section of the part before asking for the prescription.

Step by step: Designing a heat treatment recipe with AI

  1. Define target: Target hardness/toughness, in which region, according to which standard (e.g. surface hardness 55-60 HRC).
  2. Give material and cross-section: Steel type (e.g. AISI 4140), thickest cross-section, current condition.
  3. Request recipe skeleton: Ask the AI ​​for an initial recipe and justification for austenitizing temperature/time, quenching medium and tempering range.
  4. Set up an experimental matrix: Design a matrix that varies the tempering temperature in several steps (e.g. 200/400/550 °C).
  5. Apply and measure: Apply to trial pieces and measure hardness and microstructure.
  6. Verify and select: Compare results with target and standard, finalize prescription.

Experiment matrix and designed experiment (DoE)

Rather than trying each parameter one by one, a designed experiment (DoE) isolates the effects of multiple parameters with a small number of trials. AI helps in proposing a DoE matrix (e.g. full factorial or Taguchi) and interpreting the results. Below is a simple tempering matrix for AISI 4140 (austenitizing 845 °C, oil quenching constant).

trial

tempering temperature

Expected hardness trend

Anticipated satiety tendency

verification

A.

200 °C

High (~52-55 HRC)

low

HRC + Charpy

B.

400 °C

Medium (~45-48 HRC)

medium

HRC + Charpy

C.

550 °C

Low (~34-38 HRC)

high

HRC + Charpy

D.

300 °C

(risk of temper embrittlement)

need checking

Charpy + microstructure

Note: The ~260-370 °C band in line D carries the risk of "tempered martensite embrittlement" in some steels; AI can recall this, but actual behavior must be verified by Charpy testing. The hardness ranges in the table are typical trends, not exact values.

three mini cases

Case 1 — Mass effect was ignored. A workshop applies AI's prescription of "845 °C austenite, quench in oil, expect 55 HRC" for a Ø80 mm shaft made of 4140. The surface comes out to 54 HRC, but the center is only 32 HRC; because in the thick section, oil quenching cannot provide sufficient cooling speed to the center. Since AI did not ask for the section thickness and the workshop did not provide it, the prescription was incomplete. The correct approach is to give the cross-section and consider a higher hardenability steel (e.g. 4340) or harder quench media and look at the Jominy data.

Case 2 — Right matrix, right choice. A team requires both adequate hardness and acceptable toughness for a cutter. It asks AI for a matrix with three tempering temperatures as above. A/B/C is applied to trial pieces; It is seen that 400 °C gives 46 HRC and sufficient Charpy toughness. The team selects recipe B and confirms tempered martensite in the microstructure. AI designed the matrix, decision was made with test data.

Case 3 — Temper embrittlement caught. An engineer asks AI 300 °C tempering for low cost; AI states that this band poses a risk of temper embrittlement in some alloy steels and should be checked with Charpy. The engineer compares 300 °C and 450 °C; He sees that the impact toughness is lower than expected at 300 °C and increases it to 450 °C. AI pre-flagged a risk, experiment confirmed. Lesson: Even the warning of the AI ​​is tested by experiment, but the warning ensures the right experiment is set up.

Copiable prompt templates

HEAT TREATMENT RECIPE SKELETON TEMPLATE"Role: You are an assistant heat treatment engineer.Steel: [e.g. AISI 4140]. Thickest section: [mm]. Current condition: [...].Target: [surface hardness HRC / toughness / zone].Give me a START recipe and justification for austenitizing, quenching and tempering. Section thickness (mass effect) and EXPLAIN how hardenability will change the result. This is a hypothesis; should be verified by hardness/microstructure in the test piece — emphasize this.”

MATRIX OF EXPERIMENTS (DoE) TEMPLATE "I want to optimize the following parameters: [tempering temperature, time...]. Target outputs: [hardness, toughness]. Suggest me an experimental matrix (factorial/Taguchi) that separates the effect with few trials. Write down which measurement (HRC, Charpy, microstructure) I will make in each trial. Mark risky areas (e.g. temper brittleness band)."

JOMINY / HARDENABILITY COMMENT TEMPLATE "I have the following Jominy data: [distance-hardness pairs]. Comment on the hardenability of this steel: how hardness decreases with distance, up to what section thickness can I maintain the target hardness? How does it affect the choice of quenching medium? Remind me to verify the exact limit by experiment."

CCT/TTT COMMENT TEMPLATE"Comment the CCT curve for the following steel: [steel/composition]. Target microstructure: [martensite / bainite]. Roughly what cooling rate is required to obtain this microstructure and what quenching medium will provide it? State that the actual location of the curve depends on the exact composition and should be verified with my batch's curve."

Weak prompt / Strong prompt

WEAK PROMPT: "How do I harden 4140 steel?"

STRONG PROMPT:"Role: You are assistant heat treatment engineer. Part: AISI 4140, thickest section Ø80 mm, gear shaft. Target: 50-54 HRC on the surface, core hardness as high as possible at the center, good toughness. Give me the initial recipe for austenitizing + quenching + tempering, but how to reduce the center hardness of MASS EFFECT in the Ø80 mm section "Explain. If necessary, suggest an alternative steel with higher hardenability. Write down the measurements with which I will verify the recipe on the trial piece. Do not guarantee exact hardness."

The weak prompt produces a general response that is unaware of the context and target. The powerful prompt gives the cross-section and target, clearly questions mass effect, asks for alternatives, and requires verification measurements — bringing the paper recipe closer to actual furnace condition.

Common mistakes

  • Requesting a prescription without giving the section thickness (mass effect); moving the thin sample value to the thick part.
  • Mistaking the CCT/TTT "memorized" curve as the real curve of your own party.
  • Selecting the tempering embrittlement band (~260-370 °C in some steels) without checking it.
  • Measuring hardness but never measuring toughness (Charpy); the two change in opposite directions.
  • Putting the AI ​​recipe into mass production without validating it on a trial piece.
  • Not taking into account the reality of the oven (temperature uniformity, actual cooling environment).

In summary

In heat treatment, AI builds the recipe framework based on the target feature, designs the matrix of experiments (DoE) and interprets the Jominy/CCT/TTT data. However, the actual result depends on section thickness, hardenability and oven conditions; Each recipe cannot be considered reliable until it is verified by hardness and microstructure (and Charpy toughness if necessary) in the trial piece. Always ask about the mass effect, measure hardness and toughness together, and check risky temperature bands by experiment.

Application task

Choose a steel (e.g. AISI 4140 or 4340) and a realistic section thickness. With the "HEAT TREATMENT RECIPE SKELETON" template, get a recipe for the target hardness from the AI ​​and ask it to explain the mass effect. Then set up a matrix with three tempering temperatures with the "EXPERIMENT MATRIX" template, write down which measurement you will make in each trial and mark the band at risk of temper embrittlement. Finally, state in one paragraph the standard against which you will verify the recipe (hardness and toughness test).

checklist

  • [ ] I defined the target hardness/toughness, region and standard.
  • [ ] I gave the thickest section and took the mass effect into account.
  • [ ] I questioned the logic of hardenability (Jominy) and CCT/TTT.
  • [ ] I set up an experiment matrix and planned the measurement of each experiment.
  • [ ] I marked risky temperature bands (temper brittleness).
  • [ ] I confirmed the recipe with hardness and toughness in the trial piece, then put it into mass production.
Previous Next