Project overview
The supplier operated four plants with a mix of modern PLC-controlled lines and legacy equipment. Downtime was tracked on whiteboards and maintenance was reactive.
CodeQube built an edge-to-cloud IoT platform on Azure with real-time OEE dashboards, anomaly detection and predictive maintenance models integrated with the client’s SAP PM module.
Business problem
Where the client started.
No consolidated view of line performance; OEE calculated manually and inconsistently across plants.
Unplanned downtime on critical presses cost an estimated $40k per hour.
Legacy machines had no network connectivity or standard protocols.
Maintenance work orders were raised after failures, never before.
Proposed solution
What we proposed, and why.
Edge gateways running Azure IoT Edge to normalise OPC-UA, Modbus and retrofitted sensor data.
A streaming pipeline computing OEE, cycle-time and quality metrics in real time per line.
Anomaly detection and remaining-useful-life models for the twelve most critical asset types.
Automatic SAP PM work-order creation when models predicted failure within a maintenance window.
Architecture
How the system is put together.
Edge gateways buffer and normalise machine data, forwarding it to Azure IoT Hub. Stream Analytics computes live metrics into a time-series store; batch jobs train and score ML models, and results surface through a React dashboard and SAP integration.
Technology stack
Backend, frontend, data and cloud.
Challenges faced
What got in the way, and how we dealt with it.
Challenge
A third of the estate predated networked controllers.
Resolution
Retrofitted industrial sensors and current clamps feeding edge gateways; inferred machine states from power signatures where no PLC data existed.
Implementation details
How delivery was sequenced.
- 1
Pilot line (months 1–3)
Connected one press line end-to-end, validated OEE calculations with plant engineers and proved the edge architecture.
- 2
Plant 1 rollout (months 4–6)
Scaled to 300 machines, delivered dashboards to shop-floor displays and shift-handover reports.
- 3
Predictive maintenance (months 5–9)
Anomaly detection live on critical assets, SAP PM integration and maintenance feedback loop.
- 4
Multi-plant scale (months 9–11)
Remaining three plants onboarded using templated edge deployments and IaC.
Non-functional requirements
Security and scalability considerations.
Security
- IEC 62443-aligned segmentation between OT and IT networks with brokered connectivity through a DMZ.
- Device identity via X.509 certificates managed by Azure Device Provisioning Service.
- Role-based dashboards; no write path from cloud to machine controllers.
Scalability
- Templated gateway images and Terraform modules onboard a new plant in under two weeks.
- Ingestion tested to 50,000 messages/second across the four-plant estate.
- Time-series retention tiers keep hot query performance while archiving history to the data lake.
Results & business impact
What it delivered.
Business impact
- Predictive maintenance paid back platform investment within the first year.
- Shift-handover reporting eliminated manual OEE compilation across all plants.
- Platform reused for energy monitoring, cutting electricity consumption by 9%.
More case studies