A hot-bending trial rarely fails on one obvious fault. It fails as a collection of small deviations: a bend that springs back a fraction of a millimetre, a tool mark on the concave face, a batch where four chambers drift apart. Before the machine is accepted, the test plan should say what will be recorded, how it will be recorded and what the acceptance limit is. The list below follows the parameters published on our 3D curved glass hot bending forming machine page and shows what each reading is meant to prove.
1. Cycle time and throughput on a normal cycle
The equipment is specified at a cycle time under 180 seconds per piece, with 30 molds in the normal cycle. Record the actual cycle over consecutive pieces in one uninterrupted run, not the best single piece. Log the time from load to unload for each chamber, because an average can hide one chamber running slower than the others while the line total still looks acceptable. Note also where the operator is waiting: a cycle time that only holds with preheated molds and pre-loaded glass is not the cycle time you will get on a working shift.
2. Applicable size and the geometry you actually run
The applicable size on the page is 14 inch within. The trial should therefore use your own part geometry, not a simple flat coupon. Record the curved surface you need to hold — radius, rise, edge profile — and measure the formed part against the drawing after cooling, not while it is still on the mold. Springback is the number that decides whether the mold is finished, so a measurement method needs to be agreed in advance: contact probing, optical measurement or a gauge fixture, with the same fixture used through the trial.
3. Forming temperature and heating mode
Forming temperature is listed at 600–850 °C, with dual high-frequency heating as the working mode. Capture three things: the set value on the touch screen, the measured value where it matters, and the time from start of heating to forming. A controller can display a stable setpoint while the glass surface differs across the part. Where your process is temperature-sensitive, run the same recipe twice and compare the resulting geometry before you tune anything. Heating time and dwell are inputs you will later use to explain a change in cycle time, so they belong in the same record as the cycle measurement.
4. Chamber oxygen content and mold atmosphere
The machine performs circulating oxygen detection on four nitrogen-filled chambers, with a target cavity oxygen content of ≤50 ppm, and the controller value for each chamber can be set. Log each chamber separately at the same point in the cycle: after purge, at load, at forming and after unload. Four chambers that all sit well under the limit at the start of a run can diverge as the run continues, and that divergence shows up first as a batch-to-batch difference in surface quality or yield. Keep the purge sequence and purge duration with these readings — an oxygen figure without the purge settings behind it is not repeatable.
5. Product pass rate and failure rate over a batch
The published figures are a product pass rate of ≥80% and an equipment failure rate of ≤2%. Both are batch statistics, so a ten-piece sample proves very little. Run a full normal-cycle batch, with the 30-mold normal cycle as the unit, and classify every reject by cause: shape out of tolerance, surface defect, breakage, contamination, or a machine interruption. That classification matters more than the headline percentage, because a pass rate sitting on the limit for shape reasons and one sitting on the limit for breakage need completely different corrections. Record which chamber and which mold position each defect came from.
6. Mold condition and thickness monitoring
The machine detects and monitors the thickness of the mold before discharging. Treat that reading as a trend, not a pass/fail on day one. Save the value from the first accepted run as the reference, then log it at fixed intervals across the trial so wear becomes visible as a curve rather than as a surprise after months of production. Alongside the thickness reading, keep a dated photograph of the mold surface and a note on any cleaning or coating done between runs — a forming defect that appears gradually is usually a mold-surface story.
7. Recipe setting, display and traceability
Heating temperature, heating time and similar parameters can be set and displayed on the touch screen. During the trial, keep a copy of every recipe that produced an accepted part: parameter values, the chamber used, and the result. This is what lets a good run be reproduced on the next order, and it is what an engineer needs when a line is asked to run a new part geometry later. A trial that produces one good part with no saved parameters has produced an anecdote, not a process.
8. What to bring to the trial
Three items make the difference between a demonstration and an acceptance test: your own drawing with the tolerances you will hold in production, a measurement method fixed in advance, and a data sheet that is filled in run by run rather than remembered afterwards. If you already know the part family you intend to produce, send that geometry with the target cycle time and the pass rate you need, and we will set the test against those numbers instead of a generic sample.
9. Acceptance criteria agreed before, not after
Write down the limit for each item above before the trial starts: cycle time measured on which run, geometry tolerance on which fixture, oxygen content per chamber at which point in the cycle, pass rate over how many pieces. Agreement on those numbers is what allows a machine to be signed off on data rather than on impressions. Where a figure sits between the published specification and your own production target, say so before the trial so the test can be extended to cover it.
Shenzhen PowerDe Automation Precision Equipment builds this equipment alongside our semiconductor process equipment range, and process parameters on the machine can be adjusted to the glass and part geometry you are working with. Send your drawing, target cycle time and required yield to our team, and we will confirm which parameters can be reached on your geometry before any trial is scheduled.
