Machine data collection — CNC, legacy machines, Industrial Edge

Before any indicator exists, the machine has to start talking. We build the data collection layer for the whole plant — new controllers, CNC machines and machines older than some of the staff — so that all of them report in one format. The work is delivered with our manufacturing partner in Türkiye and settled with a Polish company.

Connecting older machines

A machine with no controller, or with a closed control system, still shows everything that matters on the outside. The cycle is visible on the main drive contactor, the running state on the signal tower, pieces on an inductive or optical sensor placed on the working motion, and the difference between running and idling is shown most honestly by a current measurement through a transformer. Add a small panel or a button for closing downtime reasons if the machine is to appear in the report on equal terms with the rest.

One rule holds throughout: we do not interfere with the machine's safety circuits or control logic. Signals are taken in parallel, through separated inputs, so adding measurement does not change how the machine runs and does not trigger a new conformity assessment. If the machine is still under the manufacturer's warranty, we agree the installation scope with them before starting.

CNC controls — Sinumerik and the rest

From machine tools we collect what a piece counter can never reconstruct: the name and version of the running program, cycle time, spindle and feed status, the overrides an operator applies, control alarms and tool running time. Sinumerik controls expose this through an OPC UA server, so connecting them requires no intervention in the control system. On older machines without that option we fall back on the signals described above and add data from the controller that handles the periphery.

Controls from other vendors — Fanuc, Heidenhain, Mitsubishi — are checked for the specific machine: the control version and the communication options actually purchased decide what is available. We record the result machine by machine in the offer instead of promising support for a whole family.

Siemens Industrial Edge and IIoT gateways

Data needs somewhere to pause on the way. The edge layer is built on Siemens Industrial Edge devices or on industrial IIoT gateways, DIN-rail mounted and 24 V DC powered, inside the existing cabinet or in a separate box at the machine. The edge device reads the controller over S7 or OPC UA, turns signals into events, buffers them locally and publishes them onward — usually over MQTT — to the production monitoring system or a plant database.

The buffer matters more than it sounds. A dropped link or a server restart must not erase a shift's worth of machine hours, and across several dozen machines that happens regularly. Edge applications are updated centrally, so adding another machine does not mean a walk around the shop with a laptop. The protocols we read most often are Simatic S7, OPC UA and Modbus TCP; anything else is confirmed at offer stage. If a machine needs a new controller or a rebuilt cabinet along the way, we are already in the scope of production line modernisation.

OT network security and remote access

Collecting data always raises the network question. We design this layer so that it does not enlarge the attack surface: the control network separated from the office network, traffic passing through defined points, connections initiated from the plant outwards, no ports open to the internet. Service access is personal, switched on by an authorised person on your side for the duration of a task, and logged together with what was done. The zone and conduit model described in IEC 62443 is our design reference; we do not run certification audits and we do not sell a certificate.

There is a formal side to it as well. The amended Polish act on the national cybersecurity system, which implements the NIS2 directive, places obligations on part of the manufacturing sector — if your company is in scope, every remote connection to a machine has to be describable, limited and demonstrable during an inspection. It is cheaper to design that while connecting the data than to explain it later. The rules we work to are collected in the knowledge base.

What the data is then used for

The data layer improves nothing on its own — it is the base for four uses that usually arrive in this order. OEE and downtime: availability, performance and quality calculated from signals, reasons closed at the machine; that is the scope of the production monitoring and OEE page. Maintenance: MTBF and MTTR per machine and per cell, with services planned on actual running time instead of the calendar. Energy: meters with a Modbus output on the same layer make energy per piece or per tonne calculable, and show what idling costs. Traceability: batch, order and machine settings tied together, which is what you need for a complaint or a customer audit.

Pricing

We price per machine, because machines differ: a controller with a ready interface costs nothing like a press that needs added sensors and a box. The machine survey asks for the type and year of the control system, the signals available, the distance to the nearest network point, the conditions at the mounting location and the window in which work can be done. The offer that follows separates hardware, installation and configuration. We usually start with a handful of machines — enough to test the assumptions before you price the whole shop.