Home > Docs > Industries > Manufacturing

⚙️ Manufacturing — IoT Telemetry & Predictive Maintenance¶

Last Updated: 2026-05-05 | Version: 1.0.0

"Unplanned downtime costs industrial manufacturers an estimated $50 billion per year — predictive maintenance powered by streaming telemetry can cut that figure in half."

📑 Table of Contents¶

• Scenario Overview
• Regulatory Landscape
• Data Flow Architecture
• Why Fabric for Manufacturing
• Getting Started
• References

🎯 Scenario Overview¶

📋 Regulatory Landscape¶

🏗️ Data Flow Architecture¶

flowchart LR
    subgraph Sources["🏭 Data Sources"]
        PLC["PLCs & SCADA<br/>(OPC UA)"]
        IOT["IoT Sensors<br/>(IoT Hub)"]
        MES["MES / ERP<br/>(SAP, Oracle)"]
        QMS["Quality Mgmt<br/>System"]
        CMMS["CMMS / Maint.<br/>Work Orders"]
    end

    subgraph Bronze["🥉 Bronze Layer"]
        B1["Sensor Telemetry<br/>(Eventstream)"]
        B2["Production Orders<br/>(CDC → Delta)"]
        B3["Quality Inspections<br/>(batch extract)"]
        B4["Maintenance Logs<br/>(append-only)"]
    end

    subgraph Silver["🥈 Silver Layer"]
        S1["Telemetry Time-Series<br/>(validated, downsampled)"]
        S2["Production Runs<br/>(reconciled)"]
        S3["Quality Measures<br/>(SPC control limits)"]
        S4["Asset Master<br/>(enriched)"]
    end

    subgraph Gold["🥇 Gold Layer"]
        G1["OEE Dashboard<br/>Star Schema"]
        G2["Predictive Maint.<br/>Feature Store"]
        G3["Quality Control<br/>Charts"]
        G4["Energy & Yield<br/>KPIs"]
    end

    subgraph BI["📊 Consumption"]
        DTB["Digital Twin<br/>Builder"]
        EVH["Eventhouse<br/>(real-time KQL)"]
        DL["Direct Lake<br/>Semantic Model"]
        PBI["Power BI<br/>Dashboards"]
        DA["Data Activator<br/>Maint. Alerts"]
    end

    PLC --> B1
    IOT --> B1
    MES --> B2
    QMS --> B3
    CMMS --> B4

    B1 --> S1
    B2 --> S2
    B3 --> S3
    B4 --> S4

    S1 --> G1
    S1 --> G2
    S2 --> G1
    S3 --> G3
    S4 --> G2

    G1 --> DL --> PBI
    G2 --> DL
    G3 --> DL
    G4 --> DL
    B1 --> EVH --> DTB
    G2 --> DA

💡 Why Fabric for Manufacturing¶

Digital Twin Builder is native to Fabric. Unlike standalone digital twin platforms that require separate infrastructure and data pipelines, Digital Twin Builder models production assets directly on top of Eventhouse — every property update is KQL-queryable, visualizable, and actionable through Data Activator without data movement.

OT and IT data converge in OneLake. PLC telemetry, MES production orders, CMMS work orders, and quality inspection records all land in a single governed data lake. No more reconciling siloed OT historians with ERP extracts.

Predictive maintenance without a data science team. AutoML model endpoints train anomaly detection and remaining-useful-life models on telemetry feature stores, scoring hourly and triggering proactive maintenance work orders through Data Activator.

SPC and quality analytics at Direct Lake speed. Control charts, Cpk/Ppk calculations, and defect Pareto dashboards run at sub-second speed via Direct Lake, giving quality engineers live visibility into production line performance.

Secure OT/IT boundary. Managed private endpoints and Outbound Access Protection ensure that factory-floor data ingested through IoT Hub never traverses public networks — critical for IEC 62443 compliance.

🚀 Getting Started¶

1. Ingest IoT telemetry — Configure Eventstreams to pull sensor data from Azure IoT Hub into Eventhouse for real-time KQL analysis.
2. Build the medallion Lakehouse — Follow Medallion Deep Dive for production, quality, and maintenance data.
3. Create digital twins — Use Digital Twin Builder to model production lines, binding live telemetry to entity properties.
4. Deploy predictive maintenance — Build Gold-layer feature tables and train anomaly models via AutoML.
5. Set up quality dashboards — Create SPC control chart reports with Direct Lake semantic models over Gold quality tables.
6. Wire maintenance alerts — Configure Data Activator to open work orders when predictive scores exceed thresholds.

📚 References¶