Platform Overview
AI Native Digital Engineering Data Pipeline
No installation needed  ·  Works directly in your browser  ·  Panel 10 suggested questions need no key
Full-stack pilot for industrial machinery & automotive. This demo covers all 9 layers of the AI-native PLM architecture — from mission CONOPS through MBSE, FMEA, CAD, embeddings, graph, digital twin, supplier risk, to GraphRAG AI reasoning.
What's new vs previous demo
NEWCONOPS & Mission — stakeholder needs, MoE, KPP
NEWMBSE System Model — functions, interfaces, parameters
NEWRequirements & FMEA — failure modes, RPN, mitigations
NEWDigital Thread — end-to-end traceability traversal
NEWDigital Twin Feed — telemetry, deviation alerts
NEWSupplier & BOM Risk — lead time, single-source flags
CAD Ingestion, Embeddings, Graph, Search, Reasoning
Demo scenario
Domain: Turbocharger compressor stage (industrial machinery) + EV battery bracket (automotive)
Situation: A new Gen5 compressor blade is under review.
Goal: AI should answer — is it safe to release? What are the risks? What does the field say? What is the supply chain exposure?

The answer requires ALL 9 layers. No single layer is sufficient alone.
The question only the full stack can answer
"The Gen5 blade will operate at 340°C — 5°C above the Gen3 design point. Evaluate release readiness: which system KPP is at risk, which FMEA failure mode activates, what does the Gen3 digital twin show at this temperature, and what is the supply chain exposure for the required thermal barrier coating?"
Stage 01 · New Layer
CONOPS & Mission NEW
The layer above requirements. Defines the operational concept — what mission the system serves, who operates it, under what conditions, and what measures of effectiveness it must achieve. Every downstream requirement, design decision, and FMEA risk traces back here.
Mission definition
Turbocharger Compressor Stage 3
DomainIndustrial machinery / Power generation
OperatorProcess plant continuous duty cycle
EnvironmentAmbient −20°C to +55°C, dusty, corrosive
RegulatoryISO 10816, ATEX Zone 2, PED 2014/68/EU
Service life80,000 hours TBO
Measures of Effectiveness (MoE)
IDMeasureTargetStatus
MoE-01Stage isentropic efficiency≥ 91%At risk
MoE-02Time between overhaul≥ 20,000hOK
MoE-03Pressure ratio stability±0.5% at rated speedOK
MoE-04Noise signature≤ 85 dB(A) at 1mOK
Key Performance Parameters (KPP) — flagged for Gen5
IDParameterLimitGen5 projected
KPP-01Blade tip temperature≤ 340°C340°C (at limit)
KPP-02Blade fatigue life≥ 10,000h8,400h (est.)
KPP-03Stage efficiency≥ 91%91.4%
CONOPS loaded — 4 MoE · 3 KPP · 2 regulatory frameworks
Stage 02 · New Layer
MBSE System Model NEW
The system model defines what the product must do (functions), how subsystems connect (interfaces), and what mathematical constraints link parameters to KPPs (parametric model). This gives every part a functional address — not just a number and geometry, but a role in the system's physics.
Block Definition (BDD)
Internal Interfaces (IBD)
Parametric Model (PAR)
Functions — click to expand. Each function links to parts that realize it and failure modes that threaten it.
Interface Definition Block — compressor stage
Parametric constraints — KPP derivation chain
ParameterSymbolEquationDrives KPPGen5 Value
Isentropic efficiencyη_s(T2s−T1)/(T2−T1)KPP-0391.4%
Tip clearance ratioτ/hδ_tip / blade_heightKPP-01, KPP-030.018
Creep life consumptionCLCΣ(t_i / t_rupture_i)KPP-020.84 (crit)
Blade loading coefficientψΔH / U²KPP-030.42
Pressure ratioπP_out / P_inMoE-033.52
Critical finding from parametric model
The Creep Life Consumption (CLC) parameter for Gen5 at 340°C is 0.84 — within 16% of rupture. This propagates to KPP-02 (blade fatigue life) at 8,400h estimated, 16% below the 10,000h threshold. This finding cannot be discovered from CAD geometry alone — it requires the parametric model.
Stage 03 · New Layer
Requirements & FMEA NEW
Requirements derived from MoE/KPP, and the FMEA risk register linked to system functions. RPN = Severity × Occurrence × Detection. Items above RPN 150 require mitigation before release. The Gen5 blade activates two high-RPN failure modes.
Requirements
FMEA Risk Register
Digital Thread
IDRequirementSourceStatusLinked FMEA
REQ-T-001Blade tip clearance ≤ 0.3mm at operating tempKPP-01VerifiedFM-001
REQ-T-007Thermal barrier coating mandatory above 325°CKPP-01 / FM-001NOT MET — Gen5FM-001
REQ-M-004Blade fatigue life ≥ 10,000h under cyclic loadMoE-02 / KPP-02At risk — 8,400h est.FM-002
REQ-S-002Safety factor ≥ 1.2× at max operating speedISO 10816VerifiedFM-003
REQ-S-020Disc burst margin ≥ 1.2× at 110% NmaxPED 2014/68VerifiedFM-004
REQ-P-003Stage efficiency ≥ 91% at rated conditionsMoE-0191.4% projected
Digital thread — Gen5 blade (CB-2024-NX) end-to-end traceability
Thread integrity check
Thread linkStatusGap
Mission → MoE✓ Complete
MoE → KPP → Requirements✓ Complete
Requirements → FMEA✓ Complete
Requirements → Design (Part)⚠ PartialREQ-T-007 not realized in Gen5
Part → Simulation⚠ PendingGen5 thermal FEA not yet run
Part → Field/Twinℹ Proxy onlyUsing Gen3 twin data as surrogate
Stage 04
CAD Ingestion
Parts are selected and ingested into the knowledge graph and vector index. The pipeline extracts geometry, design features, and metadata — fusing them into a single 768-dim embedding that captures not just shape, but functional meaning through requirement linkage scores.
CAD event stream — simulated (real backend: POST /api/ingest)
00:00:00INIT Geometry pipeline ready — STEP parser · PointNet++ geometry · GNN feature extractor · sentence-transformer
00:00:01WAIT Listening for STEP file uploads on POST /api/ingest...
Stage 05
Embedding Engine
Each part produces three embeddings: geometry (PointNet++ on point cloud), features (GNN on CAD feature tree), and metadata + requirements text (sentence-transformers). All three are fused into a 768-dim vector stored in Qdrant. The requirements linkage is new — it means embeddings carry functional meaning, not just shape.
← Ingest parts first in Stage 04
Stage 06
Product Knowledge Graph
The full product knowledge graph — MBSE functions, requirements, FMEA failure modes, parts, simulations, digital twin units and suppliers — all as connected nodes. Hover any node to inspect. The digital thread is traversable in both directions: from mission down to a part, and from field failure back up to the MoE it violated.
● Parts / Assembly ● System functions ● Requirements ● Simulations ● Failure modes ● Mission / MoE
Click any node to inspect
Select a node to view its full relationship context and thread position.
Stage 07 · New Layer
Digital Twin Feed NEW
Live telemetry from deployed Gen3 blade units (surrogate data for Gen5 evaluation). Sensor streams show blade tip temperature and vibration trending above design limits at high hours. These deviations flow back into the graph and are injected into the AI reasoning context.
Deployed fleet — Gen3 (CB-2021-C3) proxy for Gen5 assessment
UnitHoursSiteStatus
SN-284112,440hRotterdam plantNormal
SN-28477,420hAntwerp refineryTemp alert
SN-28539,100hDusseldorf plantVibe rising
SN-28613,200hLyon facilityNormal
Deviation from simulation baseline — SN-2847
Blade tip temp
+1.4%
Stage efficiency
−0.4%
Vibration RMS
+0.8%
Pressure ratio
−0.1%
⚠ Blade tip temp 341.2°C at 7,420h — 1.2°C above KPP-01 limit. Trending upward. At this rate, limit exceedance expected at ~9,000h.
Blade tip temperature trend — SN-2847 (last 72h)
AI insight from twin data
Gen3 at 335°C design point is already showing 341.2°C actuals at 7,420h — a +1.8% thermal drift not predicted by simulation. This means the simulation model under-predicts blade temperature by ~6°C at operating hours above 7,000h.

Implication for Gen5: If Gen5 is designed to 340°C limit but inherits the same +6°C drift, it will exceed KPP-01 by design at ~6,500h of service.
Stage 08 · New Layer
Supplier & BOM Risk NEW
Every part in the BOM is enriched with supplier data: lead time, single-source risk, country of origin, and risk score. This layer is particularly critical for Gen5 because REQ-T-007 mandates a thermal barrier coating from a supplier with a 6-week lead time — which directly affects release scheduling.
BOM risk analysis — compressor blade assembly
Part / ComponentSupplierLead wksSingle srcRisk
Ti-6Al-4V billet
CB-2024-NX substrate
Precision Alloys GmbH
Tier 1 · Germany
14 YES
7.2
TBC coating application
REQ-T-007 mandatory
TBC Coatings Ltd
Tier 2 · UK
6 NO
3.1
Precision machining
5-axis aerofoil profile
Aeroform CNC GmbH
Tier 2 · Germany
4 NO
2.2
Inconel 625 disc blank
IS-2022-D4 sub-assembly
Haynes International
Tier 1 · USA
18 YES
8.9
Critical path to release — supply constraints
Blocker 1: Ti-6Al-4V billet from Precision Alloys GmbH — 14-week lead, sole qualified source. Procurement must be triggered today to meet target release.

Risk 2: TBC coating — 6-week lead from TBC Coatings Ltd. Coating is not yet specified in Gen5 BOM (REQ-T-007 not met). Adding it extends critical path by 6 weeks.

Blocker 3: Inconel 625 disc (Haynes International) — 18-week lead, sole source. Highest risk item in BOM.
18
Max lead time (weeks)
2
Single-source parts
6.5
Avg supply risk score
$8,380
BOM unit cost
Stage 09
Vector Similarity Search
The Gen5 blade embedding is compared against the full parts library. Graph filters apply: only return parts with thermal simulation results AND requirement traceability to REQ-T-007. This is graph-filtered vector search — not just geometric similarity, but functionally relevant matches.
Query — CB-2024-NX · Gen5 (new design)
Compressor Blade Gen5
Ti-6Al-4V340°CNew — no simulationREQ-T-007 gap
Geometry
0.78
Features
0.65
Metadata + Req
0.82
Search parameters
Top-K5 results
MetricCosine similarity
Graph filter 1linked to thermal FEA
Graph filter 2satisfies REQ-T-007 (coating)
WeightsGeo 40% · Feat 35% · Meta+Req 25%
New vs prevReq linkage raises meta weight
Stage 10 · Full GraphRAG
GraphRAG Reasoning
The AI reasoning engine receives the complete context: CONOPS, MBSE model, FMEA register, digital thread, digital twin telemetry, supplier BOM risk, similarity search results, and the product knowledge graph. This is the full intelligence layer — answering questions that require all 9 layers simultaneously.
Context injected into the AI prompt (all 9 layers)
Mission CONOPS 4 MoE · 3 KPP MBSE functions + params 6 requirements FMEA — 5 failure modes Digital thread gaps 5 ingested parts + embeddings Product graph (31 nodes) Twin telemetry SN-2847 Supplier risk (2 single-source) Top-5 similarity matches
Suggested engineering questions
AI reasoning — full context synthesis
Click any suggested question above — answers appear instantly, no API key needed. Or type your own question.
Suggested questions work without any API key — answers are pre-reasoned across all 9 layers of demo data. Enter an API key only if you want to ask your own custom questions.