01
Working framework Markdown available Decision-support only
LMD Failure Atlas
A working vocabulary for mapping LMD failure modes to process signals, AI visibility, inspection methods, and validation evidence.
Preliminary decision-support only. Final feasibility depends on base material, geometry, service conditions, inspection requirements, and expert review.
Operating loop
Failure-evidence loop
02
Model
03
Decide
04
Verify
Failure mode
What can go wrong physically.
Process signal
What monitoring or logs might show.
AI visibility
What AI may responsibly indicate.
Validation evidence
What inspection or testing must prove.
Framework brief
Use the atlas as risk vocabulary
Preliminary decision-support only. Final feasibility depends on base material, geometry, service conditions, inspection requirements, and expert review.
Definition
A vocabulary map connecting LMD failure modes to possible process signals, AI visibility, validation evidence, and decision actions.
Problem it solves
Failure language gets messy when process observations, AI flags, and confirmed defects are mixed together.
Who it is for
Engineers, AI agents, students, and developers who need safer language for LMD risk and evidence.
Inputs required
Known material, geometry, process route, monitoring signals, suspected failure mode, inspection options, and service risk.
Output
A risk vocabulary and a recommended evidence direction, not a pass/fail result.
Evidence needed
NDT, CT, metallography, hardness, chemistry, dimensional inspection, destructive tests, or qualified review depending on the mode.
Limitations
It cannot confirm defects from AI output alone or approve final acceptance.
Example use case
Separating a visual anomaly from confirmed porosity, lack of fusion, cracking, dilution, distortion, or property mismatch.
Thesis connection
Sense signals, model failure risk, decide evidence routing, and verify with inspection.
Atlas
Failure modes, signals, AI visibility, and evidence
Porosity
Possible process signals: Melt-pool instability, shielding disturbance, feed inconsistency, unusual brightness or plume behavior.
AI visibility: May flag anomaly clusters or signal drift, but cannot prove pore size, position, or acceptance.
Validation evidence: CT, suitable NDT, metallography, density evidence, or test coupons depending on risk.
Decision action: Treat AI as inspection prioritization, not release proof.
Lack of fusion
Possible process signals: Low energy input, poor overlap, surface contamination, path or standoff inconsistency.
AI visibility: Can detect parameter/signal combinations associated with risk when trained against validation data.
Validation evidence: CT, metallography, destructive cross-section, or qualified NDT method.
Decision action: Require stronger evidence for load-bearing or fatigue-sensitive use.
Cracking
Possible process signals: Material mismatch, rapid thermal cycles, high residual stress, heat-sensitive base material.
AI visibility: May flag thermal histories or acoustic/visual patterns, but should not decide acceptability alone.
Validation evidence: PT, MT where applicable, metallography, hardness, residual-stress-aware review.
Decision action: Escalate material compatibility and preheat/postheat review.
Excess dilution
Possible process signals: High heat input, slow travel, bead geometry change, unexpected melt-pool size.
AI visibility: Can track process windows and bead-shape indicators if linked to chemistry or cross-sections.
Validation evidence: Metallography, chemistry, hardness profile, deposit/base interface review.
Decision action: Review process route when functional surface properties matter.
Distortion
Possible process signals: Heat accumulation, thin geometry, poor fixturing, long deposition time, sensitive tolerances.
AI visibility: Can forecast risk from geometry, heat input, and prior comparable jobs.
Validation evidence: Dimensional inspection, scan-to-CAD comparison, fixture and machining review.
Decision action: Plan machining allowance and in-process checks early.
Surface or bead geometry defect
Possible process signals: Track height variation, spatter, feed interruption, poor overlap, local access constraints.
AI visibility: Strong candidate for visual/process signal detection and triage.
Validation evidence: Visual inspection, 3D scan, dimensional inspection, machining allowance review.
Decision action: Separate cosmetic/process observations from final tolerance evidence.
Property mismatch
Possible process signals: Wrong feedstock, unknown base material, heat treatment gap, dilution or hardness shift.
AI visibility: Can identify missing traceability and incompatible requirements, not certify properties.
Validation evidence: Material certificates, hardness testing, metallography, chemistry, mechanical testing.
Decision action: Stop firm recommendations if material grade or feedstock traceability is unknown.
How to use it
Use the atlas as a risk vocabulary, not a pass/fail engine
The atlas is strongest when paired with the LMD Quality Evidence Ladder and RFQ schema. It helps name the risk and request the right evidence before a claim becomes too confident.
Framework path
Move from framework to verifiable decisions
Industrial AI ThesisThe shared operating model for the public frameworks.Open →Manish Sharma - Industrial AI & Decision SystemsCanonical identity page for the author and public entity.Open →LMD / DED Domain HubThe established public proof domain.Open →Full Site MapEvery page and public machine-readable asset in one linked map.Open →