Manufacturing: from logbooks to live plants

For most of the 20th century, manufacturing ran on tribal knowledge. The senior fitter knew which spindle was about to fail. The shift supervisor wrote OEE numbers on a whiteboard at end of day. Quality issues surfaced in customer complaints, not control charts.

PLCs eventually showed up on lines, but they were silos. Each vendor — Siemens, Allen-Bradley, Mitsubishi — spoke its own dialect. Data lived inside the controller. If you wanted to see it on a manager's desk, you copied it by hand.

Today, most modern plants run on a unified namespace. OPC UA standardises the conversation between PLCs, MES, SCADA, and the cloud. MQTT (often Sparkplug B) carries the data with sub-second latency. Every machine has an IP address.

The shift is from "we measure what we can" to "we capture everything, then decide what matters". OEE updates in real time. A line can be paused remotely. A quality fault on machine 14 can flag inventory at the receiving dock the same minute.

The factory of 2035 will be a closed-loop system. Edge AI will run inference on the PLC itself — vibration models, anomaly detection, predictive scheduling — without round-tripping to the cloud. Digital twins will let teams simulate line changes hours before the real machinery moves.

Industry 5.0 puts the human back at the centre. Cobots that share workspace, wearables that prevent injury, and AI that augments operators instead of replacing them. The plants that survive will be the plants that listen — to their machines, their workers, and the world outside.