Views: 0 Author: Site Editor Publish Time: 2026-11-17 Origin: Site
November 2026 — The core deployment of our Manufacturing Execution System (MES) Phase I is complete, and the project has now entered integration testing. It is not yet live.
For a foundry, the difficulty with MES is never buying the software. It is getting the system onto the shop floor. The equipment is old, the process is alive, and the operators work standing up — a single click in the software may require scanning, cross-checking and confirming three actions in the plant. So Phase I is not aiming for blanket rollout. We are connecting the backbone from melting through warehousing, commissioning station by station: validate one segment, stabilise it, then move to the next.
Why a foundry has to be able to trace. Quality problems in iron castings share one pattern: they are found late, and the root cause sits upstream. A casting shows shrinkage, gas porosity, slag inclusion or a dimensional deviation; inspection can only mark it non-conforming. The actual cause was set hours or days earlier — a chemical composition drift in a heat, insufficient spheroidisation temperature, inadequate mould compaction, pouring too fast, a heat-treatment curve that deviated from setpoint. Reconstructing that from paper records usually gets you to "probably that batch" and no further.
EN 877 regimes and customer audits are increasingly unforgiving: overseas buyers conducting factory audits ask directly for the complete production parameters behind a given batch — heat number, chemical composition, pouring temperature, heat-treatment records, inspection reports. If you cannot produce them, the order is at risk. MES exists to close that broken chain.
What Phase I is designed to cover. First, heat genealogy: every heat receives a unique number at melting, recording charge structure, chemical analysis (C, Si, Mn, P, S and alloying elements), tapping temperature, inoculation/spheroidisation parameters and pouring times. The heat number binds to each mould poured, and each mould to each casting after shakeout — so any finished part can be traced back to its heat, that heat's chemistry, the pouring temperature, the operator and the machine.
Second, process-data capture: critical stations are designed to connect to equipment via PLC and OPC UA, collecting pouring temperature and cadence, mould temperature and heat-treatment curves automatically, replacing manual log sheets, with real-time alarms when parameters leave the process window.
Third, quality closure and SPC: inspection results — radiographic or ultrasonic defect type and location, CMM dimensions, hardness and metallography — bind to the individual part number, with defects coded and trended by heat, shift, pattern and operator. Once live, shipments can carry an auditable electronic batch record rather than a handwritten certificate.
Fourth, OEE and energy visibility: uptime, downtime reasons and rate attainment for melting furnaces, heat-treatment furnaces and machining equipment. Foundry losses hide in waiting for iron, moulds and crane time; structured downtime coding is what makes improvement possible.
What testing is actually verifying. The time in integration testing goes not to software features but to four things. Do the data points reconcile — does the pouring temperature read from the PLC match the operator's pyrometer within tolerance, is the furnace sampling rate sufficient to capture the ramp, because inaccurate sources invalidate every downstream analysis. Can legacy equipment be connected — tiered by OPC UA where protocols exist, added sensors where they do not, and enforced terminal entry with error-proofing where neither is possible. Is the scanning flow usable — where the label sits, when it is scanned, whether a wrong scan can be reversed, how offline entries are backfilled; get this wrong and the floor routes around the system, so the rule is "scan, don't type." And are the transfer gates calibrated — no heat into pouring without composition confirmation, no transfer without pouring temperature, no receipt without bound inspection results; too loose and the data is dirty, too tight and the line stops.
What it will mean for export. Once running reliably: factory audits and customer qualification can pull batch genealogy, process records and inspection reports by order number in one step; and origin verification, quality claims or trade-remedy questionnaires can be answered with verifiable primary production records rather than retrospective explanation.
Next steps. After integration testing passes, Phase I moves to limited trial running on real work orders, monitoring data completeness and operator acceptance before widening coverage. Phase II will connect the order–work order–inventory loop with ERP, finished-goods batch linkage with WMS, and product carbon-footprint accounting built on actual energy data — preparation for EU CBAM-type requirements and a question customers increasingly ask.
Testing is the least visible stage of MES, but it determines whether the system ends up as software or as shop-floor habit. We would rather spend the time on site than push a system live before it runs properly.