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 10 / 12

Characterization Data Interpretation: XRD, SEM/EDS, DSC and Mechanical Testing

Gains:

  • Ability to interpret characterization data such as XRD, SEM/EDS, DSC and tensile test in a structured format with AI
  • Ability to recognize the limits and risks of artifacts in AI in peak, phase, composition and transformation temperature inferences
  • Ability to verify AI's characterization interpretation with reference database, standard and repeat measurement

You cannot see "with the eye" which phases are in a material, its chemical composition, when it melts and transforms, and how much charge it carries; characterization techniques measure these. In this unit, you will learn how to interpret the data of four basic techniques with the support of artificial intelligence: hardness, impact). AI is powerful in structuring this data, interpreting peaks/transitions, and generating reports. But artifacts, overlapping signals, and calibration problems can lead to incorrect conclusions; Each interpretation should be verified by reference database, standard and repeat measurement.

The four languages ​​of techniques and the role of AI

  • XRD: X-rays diffracted from crystal planes give peaks at certain angles. Peak positions convey crystal structure, intensities convey phase ratio, and widths convey grain/stress information. Phase identification is done by matching peaks to reference patterns (ICDD/PDF cards).
  • SEM/EDS: SEM gives surface/morphology at high magnification; EDS identifies the elements at that point (with characteristic X-ray energy). EDS is semi-quantitative; The margin of error is high for light elements and overlapping peaks.
  • DSC: The difference in heat flux between the sample and reference is measured; Melting (endothermic) and crystallization/transformation (exothermic) peaks give temperature and energy.
  • Mechanical test: Tensile test determines yield/tensile strength and elongation; hardness local strength; Charpy gives impact toughness. Each of them is made according to the standard (ISO 6892, ISO 148, etc.).

AI is helpful in converting this data into an organized table, interpreting peaks/transitions and catching discrepancies; but “reading” the raw data and final assignment requires physical verification.

Caution: Claiming "exact composition" with EDS is dangerous. EDS is semi-quantitative; Surface roughness, overlapping peaks (e.g. S and Mo, Ti and N), charging and sample tilt distort the result. Accurate composition requires WDS or wet chemistry/spectrometer calibrated with standard.

Step by step: interpreting a characterization data with AI

  1. Give measurement condition: Device, parameter (scanning speed, heating rate, X-ray source), sample preparation.
  2. Structure the raw data: Ask the AI ​​to convert the peak/transition table to regular format (position, intensity, area).
  3. Request preliminary comments: Possible phases/elements/transitions and alternative explanations (could it be artifact?).
  4. Match reference: PDF card for XRD, expected composition for EDS, literature temperature for DSC.
  5. Artifact/inconsistency addressed: Aliasing, background, preferential orientation, baseline shift.
  6. Repeat/cross-validate: Second measurement, second technique (e.g. XRD + microstructure), standard method.

Technical-inference-artifact table

technical

Main takeaway

Typical artifact/trap

verification

XRD

Phase, crystal structure

Preferential orientation, overlapping peak, background

PDF card matching, repeat

SEM/EDS

Morphology + element

Loading, overlapping peak, surface roughness

WDS/spectrometer, standard

DSC

Transition/melting temperature

Heating rate effect, baseline

Research, literature

tensile test

Yield/tensile, elongation

Slip/grip error, speed

Standard (ISO 6892), repeat

three mini cases

Case 1 — Wrong phase from single peak. A student shows a single high peak in the XRD pattern to the AI ​​and asks "which phase is this?" AI gives a phase name. However, that peak position appears at a similar angle in two different phases (overlap) and the preferential orientation in the sample distorted the intensities. When the student matches the pattern to the PDF cards as an exact pattern (all peaks), he sees that the actual phase is different. Lesson: phase identification is not made from a single peak, but from matching the entire set of peaks with the reference.

Case 2 — EDS was considered certain. A team measures an inclusion with EDS; The AI ​​interprets the output as "contains 2.1% S." However, the S peak overlaps the Mo peak and there is Mo in the sample; actual sulfur is much lower. When the team measures again with WDS and resolves the overlap, they see the difference. Lesson: EDS is semi-quantitative and overlapping peaks disprove the claim of definitive composition; The critical value is verified by another technique.

Case 3 — Correct DSC interpretation. Aging (precipitation) behavior in an aluminum alloy is studied. An exothermic peak is seen in DSC; The team has the AI ​​interpret the measurement condition. AI indicates that the peak may correspond to precipitation formation, but the heating rate shifts the peak temperature and should be rescanned at different speeds. The team measures at two different heating rates, confirms the peak, and compares it with the literature. AI gave the correct interpretation and the correct warning, and the re-measurement was finalized.

Copiable prompt templates

XRD INTERPRETATION TEMPLATE"Role: You are the characterization assistant.

EDS EVALUATION TEMPLATE"My EDS result: [element-percentage list], sample [description]. Interpret these values, but remember that EDS is SEMI-QUANTITATIVE. Mark which elements are at risk of peak overlap (e.g. S/Mo, Ti/N) or slight elemental error. Tell me with which technique (WDS,spectrometer) I should confirm for the exact composition."

DSC INTERPRETATION TEMPLATE "My DSC curve: [peak temperatures, endo/exo, heating rate]. Sample: [...]. Interpret each peak with a possible transition (melting, crystallization, precipitation), but remind that heating rate shifts the peak temperature and the baseline effect. Indicate that repeats at different heating rates and comparison with the literature are required."

MECHANICAL TESTING COMMENT TEMPLATE"Tensile test result: yield [...], tensile [...], elongation [...] %, standard [e.g. ISO 6892], sample [...]. Interpret these values: are they reasonable for the material/condition, does it meet the specification? Ask about the possibility of grip slippage, test speed or sample error. Do not rely on a single sample and indicate the need for retesting."

Weak prompt / Strong prompt

WEAK PROMPT: "What is the phase on this XRD graph?"

STRONG PROMPT:"Role: You are the characterization assistant. "Connect to matching with PDF cards. Mark overlapping peaks and risk of preferential orientation. Do not give exact phase ratio; tell me how to quantify (e.g. Rietveld)."

A weak prompt invites a hasty response from a single spade. The powerful prompt gives the entire peak set, measurement condition and sample context; ties the assignment to reference matching, flags artifacts, and prompts for quantitative verification.

Common mistakes

  • Making phase assignment from a single peak in XRD; not matching the entire set of peaks to the reference.
  • Mistaking EDS for quantitative and omitting overlapping peaks (S/Mo, Ti/N) and light element error.
  • Considering the DSC peak temperature as the "exact" value, independent of the heating rate.
  • Accepting the mechanical test result from a single sample without questioning the grip/speed error.
  • Having the AI ​​interpret the measurement condition (device, parameter, preparation) without specifying it.
  • Being content with one technique; Not cross-validating like XRD + microstructure.

In summary

In interpreting characterization data, AI is powerful in structuring XRD/SEM-EDS/DSC/mechanical test results, pre-interpreting peaks and transitions, and reminding of the risk of artifacts. However, phase assignment requires matching of the entire peak set to the reference, EDS exact composition assertion requires WDS/spectrometer confirmation, DSC transitions require rescanning, and mechanical values ​​require standard method + repetition. Always give the measurement condition, eliminate artifacts, cross-verify with the reference database and the second technique.

Application task

Choose a characterization technique (XRD, EDS or DSC) and a real/fictional data set. Ask the AI ​​to interpret the data with the relevant template (e.g. “XRD COMMENT”) and be sure to give the measurement condition. List the artifacts/traps that the AI ​​flags (aliasing, preferential orientation, heating rate, etc.). Finally, write in one paragraph which reference database, standard or second technique you will verify this interpretation with.

checklist

  • [ ] I gave the measurement condition (device, parameter, sample preparation) to the AI.
  • [ ] I matched the XRD phase assignment to the entire set of peaks and PDF cards.
  • [ ] I considered the EDS result semi-quantitative and checked for overlapping peaks.
  • [ ] I evaluated the DSC transitions together with the heating rate effect and scanned them again.
  • [ ] I confirmed the mechanical test result by standard method and repeat measurement.
  • [ ] I cross-validated the comment with the reference database and a second technique.
Previous Next