FDE Industrial AI Deployment Skill#

Open Source: https://github.com/jaccen/FDE-Industrial-Skill

Overview#

Full-spectrum support for FDEs deploying AI & big data on industrial production lines — from scenario diagnosis to scaled deployment.

Core Workflow#

text

Scenario Diagnosis -> Data Governance -> Solution Design -> POC -> Scale-up -> Feedback Loop

Step 1: Scenario Diagnosis#

1. Read references/fde-role-model.md for FDE capability framework.
2. Apply "Pain-Data-Impact" triage: Pain (business pain), Data (sufficiency), Impact (quantifiable ROI).
3. Classify into 5 core categories — references/industrial-ai-scenarios.md.

Step 2: Data Governance & Integration#

1. Map data sources: OT (SCADA/PLC/sensors), IT (MES/ERP/PLM), ET (engineering docs).
2. Palantir-style Ontology: Objects, Links, Actions.
3. Data quality gaps: missing values, timestamp misalignment, label scarcity.
4. Pipeline: edge collection -> ETL -> feature store.

Key: Start from business decisions, not data tables.

Step 3: Solution Design#

• Visual inspection: CNN/ViT + edge GPU boxes
• Predictive maintenance: LSTM/Transformer + physics-informed features; 7-14 day window
• Process optimization: RL/Bayesian + digital twin; single process first
• Energy efficiency: regression + control optimization; baseline first
• Supply chain: graph model + demand forecast + ERP integration

Step 4: POC Deployment (Zero Week)#

Day 1-3: data audit + interviews; Day 4-7: baseline model + quick wins; Week 2-4: training + integration; Week 4-6: A/B test + operator training.

Critical: Deliver measurable quick win within 2 weeks.

Step 5: Scale-up & Feedback#

Measure ROI, generalize single -> multi -> factory-wide, FDE+FDR feedback loop.

ROI Framework#

MetricTypical RangeDefect detection improvement80-95% reductionUnplanned downtime reduction30-60% reductionYield improvement2-8% increaseEnergy savings5-15% reductionROI payback period6-18 months

Reference Guide#

NeedReferenceFDE role & skillsfde-role-model.mdScenario & algorithmindustrial-ai-scenarios.mdDeployment methodologylanding-methodology.mdCase studiescase-studies.md