An injection molding press is one of the best-instrumented machines in manufacturing. It knows its cycle time to the hundredth of a second, its melt and mold temperatures, its shot count, its cushion, and whether the last shot resembled the one before it. In most plants all of that stays on the machine, and the plant's official record of the shift is an operator run report keyed into Excel days later.
What live machine data actually captures
- Cycle time per shot, not the average someone estimated, so drift becomes visible as it starts.
- Shot and part counts by mold and by cavity, which is what makes the yield math match the floor.
- Melt temperature, cushion, and pressure, the process signature that changes before scrap appears.
- Downtime with a cause, coded at the machine in seconds instead of written as misc in a notebook.
- Mold changes, timed automatically from last good part to first good part.
Any brand, without buying new presses
The common objection is fleet age and mixture. A shop running Husky, Netstal, Arburg, and a few machines older than the maintenance manager assumes that connection means replacement. It does not. Direct PLC and sensor connections work across brands, Allen-Bradley and Rockwell, Siemens, Omron, Mitsubishi and others, over OPC UA or whatever protocol the machine speaks, with sensors added only where a controller is genuinely closed. The goal is one live screen holding every press, not a uniform fleet.
What changes once the presses report themselves
Drift gets caught in minutes. Dimensional and cosmetic rejects normally surface at quality control hours after the cycle drifted, so a full shift of parts carries the same flaw into regrind. Machine learning on live cycle data flags the pattern as it emerges and signals the press, which is the difference between an adjustment and a scrap report.
Per-cavity truth appears. Output counted at the mold hides a blocked or worn cavity for weeks. Per-cavity output and reject rates make the underperformer obvious within the hour, and the yield number finally reconciles with what the floor believes.
Quotes stop being guesses. Estimating on standard cycles and standard scrap is how a shop wins work that loses money. Actual cycles, actual scrap, and measured changeover times flowing back into quoting produce margin by mold before the quote goes out. That loop is often the fastest payback in a molding business, and it only exists once the presses report themselves.
Machine data plus the paper it does not replace
Live data does not remove the need to digitize what people observe. Operator checks, material lots, and changeover notes still need capture at the press on tablets, and they belong on the same layer as the machine stream so a scrap spike can be read against the resin lot and the shift together. That pairing of machine connections and station capture is the first phase of a Harmony deployment on plastics and rubber floors, delivered on-site in a published pilot of $15-20K one time over 4-6 weeks, with working software by week three. The wider category context sits in our production tracking software guide.
Starting without boiling the ocean
Connect one press cell first, ideally the one running the highest-volume or most troublesome mold. Stream cycles, counts, and downtime, put the live screen where the floor can see it, and let a week of real data land. Two things usually happen. The measured cycle time turns out slower than the standard everyone quotes from, and the real downtime pareto looks nothing like the notebook version. Both discoveries are worth more than the cost of the connection, and together they make the case for the rest of the fleet without anyone needing a slide deck.