Splay isn't a press problem. Check the dryer logs.

A part comes off the press with silvery streaks radiating from the gate. The process tech bumps back pressure, drops the injection speed, maybe tweaks the melt temp. The streaks fade a little. Production keeps running. The scrap bin quietly fills up over the shift.

Splay gets blamed on the press because the press is where you see it. The actual cause is usually upstream, in the dryer, and the data to prove it is sitting in a log nobody opens.

Moisture is the boring answer, and usually the right one

PlasticsToday's troubleshooting column on splay is worth reading in full, but the short version is: with hygroscopic resins (PET, PC, nylon, PBT, TPU), residual moisture flashes off during injection and shows up as splay, voids, or brittle parts. The fix isn't at the machine. It's dewpoint, drying temperature, and residence time in the hopper. PlasticsToday walks through how routinely those three get mis-set or mis-monitored on the floor.

The classic failure modes we see:

- Dewpoint sensor drifted or fouled, still reporting a happy -40°F. - Drying temperature set correctly on the controller but actual hopper temp is 20°F lower because the heater is aging. - Throughput increased on the press, so residence time in the hopper dropped below spec, and nobody recalculated. - Resin loaded from a gaylord that sat open on a humid day. Nothing in the system knows.

None of that is exotic. All of it is measurable. Most of it is already being measured.

The data gap isn't sensors, it's correlation

A modern molding cell is instrumented. Cavity pressure, melt temp, fill/pack/hold signatures, screw recovery time — Tech Briefs has a good overview of what's captured every shot and how little of it drives quality decisions. The dryer has its own telemetry: dewpoint, process temp, return air temp, sometimes throughput. The material handling system knows which silo or gaylord fed which press when.

Those three data streams almost never sit in the same place. The press data goes to a process monitoring system. The dryer data lives on the dryer's own HMI, or in a standalone file, or nowhere. The resin lot info is on a paper traveler or in a receiving spreadsheet.

So when splay shows up on Tuesday afternoon, nobody can answer the actual diagnostic question: what was the dewpoint during the two hours before this part was molded, and what lot of resin was in the hopper?

What actually moves the needle

You don't need an AI model to solve this. You need three things joined on a timestamp:

1. Dryer dewpoint and process temp, logged at 1-minute resolution. 2. Press shot data, with the defect disposition from QA. 3. Resin lot and load time from material handling.

Once those three sit in one place, splay events cluster visibly against dewpoint excursions and lot changes. That's it. That's the project. The alert that matters is "dewpoint has been above -20°F for the last 45 minutes on Dryer 4," delivered before the next shift finds the parts.

The instrumentation is already there. The decision workflow isn't. That gap is where the scrap lives.