Industrial sensors on a plant floor generate vibration, temperature, and pressure data. Alone, it’s noise. Joined with IBM i asset records, maintenance history, and production orders, it becomes predictive intelligence. This project builds that join.
The pipeline: sensors → MQTT → Node-RED → Kafka → stream processor → IBM i Db2 (via Debezium CDC for the business context side). The dashboard merges both streams in real time, surfacing alerts with full business context — which machine, which SLA, which production order is affected.
The ML step: train a failure-prediction model on historical sensor + IBM i maintenance data, deploy it as a Kafka Streams processor, write predicted maintenance work orders back to IBM i automatically. A complete closed loop.