Embedded Controllers & Industrial PCs for Plastics & Injection Moulding 2026 — Euromap 77, In-Mould Sensing & Shot-Level Traceability

Injection moulding is one of the most deterministic manufacturing processes there is. A press repeats the same cycle thousands of times a shift — close, inject, hold, cool, open, eject — and every repetition produces one shot and one set of parts whose quality was decided by a handful of variables the machine already controls and measures. That is the unusual opportunity on a moulding floor: the data needed to qualify a part is not something an operator has to go and find, it is generated by the press on every cycle and merely needs to be captured, judged and recorded.

Plastics processors are also among the most cost-constrained buyers of automation, which is why the interesting question is rarely whether to add compute but where the cheapest useful tier of it belongs. This guide covers the press-floor environment, the workloads worth running at machine level, the Euromap 77 / OPC UA data model that makes them possible, and the specification rules that decide whether a box survives oil mist, hydraulic heat and a decade of three-shift running.

The Press Floor as a Data Source

A modern moulding cell is a cluster of machines from several generations. Some presses are new, speak OPC UA natively and expose an object model; some are twenty years old and offer only a serial port or a relay contact. The first hardware decision is therefore not “which industrial PC” but “how much of the fleet can be instrumented without replacing it,” and that determines whether the project is a gateway retrofit or a controller replacement.

Euromap 77, standardised as OPC UA companion specification 40077, is the reason this is tractable. It defines a common object model for an injection-moulding machine so that a supervisory node can read the same fields from presses of different makes, rather than writing a bespoke driver per controller. Where a press predates it, Euromap 63 or plain Modbus still yields the cycle and state signals that a process monitor needs; the full field set is a convenience, not a precondition.

The Workloads and Where the Compute Sits

The moulding floor produces fast, in-cycle data at the machine and slow, aggregate data at the plant, and the two need opposite treatment. Cavity pressure resolves the melt front and holds pressure in milliseconds, so it only means something when a node samples and reduces it in the cycle; an energy counter produces a value per shift that any server can sum. The rule that keeps the architecture honest is the same one used across process industries: reduce at the machine, transmit the conclusion, and keep the plant historian for what genuinely needs history.

The clearest case where the edge decision is made by the process clock rather than by convenience is cavity pressure. A pressure trace through one cycle is a curve, not a number: it rises through injection, holds, then falls as the part cools. A part is good when the curve matches the golden curve for that mould within tolerance; a short shot, a partially blocked gate or a cold mould shows up as a distorted curve long before a human notices a bad part. For that judgement to be worth anything it has to be made as the parts are produced and out of the cycle, which is exactly a local-node workload. A press running a thirty-second cycle produces about 120 shots an hour, or roughly 960 a shift; across a forty-machine shop that is around 38,000 shot records a day before any vision data, and the useful output is still a pass/fail plus the exception, not an archive of every trace.

It is equally important to say where this must not sit. Cavity-pressure monitoring and vision are per-cycle quality decisions; they are not real-time servo control. The press's genuinely fast loops — injection velocity profiling, clamp force, ejector motion and every safety interlock — belong in the machine controller and the safety chain, and an edge node must never be inserted into their path. The correct relationship is one-way: the node reads Euromap 77 fields and sensor signals, produces quality and maintenance conclusions, and never closes a control loop.

Selection Rules for a Moulding Floor

Sizing the hardware is mostly a matter of refusing to compromise on the environment and the clock. The table below lists the specifications worth writing into a purchase order, because each one maps to a failure mode that returns within a year or two if it is skipped.

• Instrument before you replace: a gateway that reads Euromap 63/77 from the existing fleet usually beats retrofitting machines.
• Standardise on the Euromap 77 / OPC UA 40077 object model so a supervisory node is machine-maker agnostic.
• Reduce at the machine and transmit the conclusion — cavity-pressure traces in, per-shot pass/fail out.
• Keep cavity-pressure and vision judgement local, because a 1–10 kHz trace is only useful as a feature, not as a stream.
• Keep every injection, clamp and safety loop in the machine controller; the node advises and never closes a critical loop.
• Tag every record to a part or lot at the moment of production, or traceability is a reconstruction rather than a record.
• Timestamp across the plant with PTP so a cavity-pressure anomaly and a downstream defect can be correlated later.
• Version-pin firmware and keep a documented update path across a fleet that will outlive several ERP upgrades.

QSCompute supplies the compute and storage that sit behind the instruments: fanless wide-temperature industrial PCs and compact embedded controllers for machine-side nodes, edge GPU systems for take-out vision, Euromap 77 / OPC UA gateways and power-loss-protected industrial storage sized for shot-level traceability. Send us the press list, the sensor set and the data rate, and we will size the node that fits the cell it lives in.