How to Use Cavity Pressure Sensors: In-Mold Process Control

A thin-wall lid program can run 2.5 percent scrap for three weeks while the press itself looks perfectly stable. The hydraulic curves repeat within 1 percent, yet the melt front inside the cavities changes with every resin lot. The signal that catches that drift is the one most tools never carry: cavity pressure measured at the steel that forms the part.

This guide covers the sensor layer: families, tip placement, the four decision signals on each curve, mounting and wiring, closed-loop options, the cost per tip, and a retrofit checklist.

The Snapshot

• 0.5 to 1 percent repeatability. Peak cavity pressure repeats shot to shot within 0.5 to 1 percent on a stable process, against 5 to 10 percent scatter on clamp force or screw velocity curves.
• Tips rated to 2,000 bar and 250 °C. Piezoelectric flush diaphragm tips survive melt temperature, injection pressure and nozzle contact, and they read the pressure that actually packs the part.
• 10 percent peak spread is the balance alarm. Multiple tips in one mold expose runner or gate imbalance when peaks spread more than 10 percent across the cavity set.
• 600 to 1,800 USD installed per tip. A sensor costs 500 to 1,500 USD and mold-side machining adds 100 to 300 USD per tip, before the amplifier and software that turn the signal into a per-shot decision.

Why Cavity Pressure Beats Machine Hydraulics

A molding machine reports its own intent. Screw position says how far the ram traveled, hydraulic pressure says how hard it pushed, and clamp force says how tightly the tool closed. None of those numbers describes the melt at the gate. Between the screw tip and the cavity sit 100 to 300 mm of compressible polymer and millimetres of steel flex, and that column absorbs every change in the material.

Those changes arrive with every resin lot. A filled polyamide whose melt flow index drifts 1 to 3 g/10 min, or a dryer running 10 °C low, shifts melt viscosity. The press hides it: the same hydraulic curve now produces a different cavity peak. Viscosity moves peak pressure 10 to 20 percent while the machine's own curves stay unchanged.

Cavity pressure measures the deformation event itself. The diaphragm sits in the flow path, so the reading captures fill resistance, packing and freeze exactly as the part experiences them. A stable process holds its peak within 0.5 to 1 percent shot to shot, and no press curve can verify that number. Pressure does not replace the machine's controls; it audits them.

The distinction matters where tolerances are tightest. A connector housing at ±0.05 mm cannot be validated by one shot per hour. Per-shot pressure data underpins the method in our scientific molding study guide: the study sets the profile, and the sensor proves it landed on every cycle.

Sensor Families and Pressure Ratings

Three families cover injection molding use. The piezoelectric flush diaphragm tip puts a quartz element behind a polished membrane 3 to 6 mm wide, rated to 2,000 bar and melt temperatures up to 250 °C, some to 350 °C. It has almost no drift and responds in under 1 ms, which is why it dominates process control. Its charge output needs a dedicated amplifier near the mold.

Strain gauge or thick-film melt transducers read 0 to 1,000 bar, tolerate repeated purge cycles and cost less per channel. They suit gate freeze timing and trend records rather than microsecond switchover decisions. The differential tip carries two sensing faces in one bore, upstream and downstream of the gate. Subtracting them isolates the pressure drop across the gate land, and when that drop collapses the gate has frozen.

← swipe to scroll →

Table note: ranges and temperatures are typical published values for injection molding sensor hardware. Confirm the exact tip rating against the resin processing window before ordering, since a 300 °C polycarbonate run exceeds a 250 °C diaphragm.

Budget the electronics per channel, not per mold: a piezo channel needs its charge amplifier within roughly 2 m of the tip. A 16-cavity tool with one reference tip and a gate differential is a three-channel system.

Tip Placement Rules: Gate Land or Last-to-Fill

A tip within 1 to 5 mm of the gate entrance reads the pressure that drives the part: fill, packing and freeze of the whole cavity with minimal attenuation. Its blind spot is that the gate can seal while the far end of the part keeps shrinking.

A tip in the last-to-fill region answers the cosmetic and dimensional questions. It records when melt reaches the far corner, where sinks, burns and incomplete packing appear first, and on multi-cavity tools it becomes the family reference for balance. The trade-off is attenuation: flow resistance between gate and tip smooths the curve, so setpoints are learned on the instrumented mold, never copied from a datasheet.

Three placement rules protect the reading. First, keep the tip within 1 to 3 mm of the flow path with the diaphragm flush in the bore wall, so melt contacts the sensor directly; deeper steel adds lag and erases the knee that triggers switchover. Second, never place a tip in dead steel: a blind corner or vent land reads compressed air or nothing at all. Third, keep the tip out of the weld line path and out of the jetting stream 5 to 10 mm ahead of an open gate, where an unmelted snake strikes the diaphragm before the cavity fills.

For a first tip on a single-cavity tool, choose the gate; on a multi-cavity tool, add the last-to-fill reference. Those two points expose fill, pack, freeze and imbalance. Decide placement during DFM, while the bore is still a drawing line, because deciding later turns a design choice into a repair, as our mold trial guide covers for every other feature on the tool.

The Four Decision Signals on a Pressure Curve

The first signal is the fill switch-over at the viscosity knee. The trace climbs gently as the front sweeps the cavity, then breaks hard upward when the last region fills and compression replaces flow. That knee, typically 200 to 400 bar at the reference tip, marks the moment to leave velocity control for pack. Triggering on the cavity signal instead of a screw position preset removes cushion scatter from lot viscosity swings.

The second signal is the peak pressure window. On a stable process the peak repeats within 0.5 to 1 percent, so a practical alarm band is ±5 to 10 percent of the qualified value. Below it the shot is short of packing; above it, flash and gate drag threaten the tool. The peak also feeds clamp sizing, since cavity pressure multiplied by projected area is the separating force that the clamp tonnage calculation must survive with margin.

The third signal is gate freeze, read straight off the decay curve. After pack, pressure falls as the part shrinks and the gate solidifies, flattening when the gate seals, often below 50 bar of residual drop across the land. A differential tip makes the inflection explicit, and ending hold at true freeze removes the seconds a guessed timer pads onto every cycle.

The fourth signal is the ejection condition. A part that sticks shows a sub-atmospheric dip or a sharp spike in the final 200 ms, and the trace shape says which. A weight check never catches it, and vision catches it only after the part tears.

Qualify every window on the actual tool during T1 sampling, because tip position and gate size shift the absolute numbers.

Cavity Balance Diagnosis From Multiple Tips

One pressure curve proves one cavity. Family molds need a reference tip per cavity, or at minimum one in the suspected fast cavity and one in the slow cavity, to make balance a measurement instead of an inference. The acceptance number is simple: when peaks spread more than 10 percent across the cavity set, the runner or gate system is unbalanced, and the parts will not weigh alike either.

The spread locates the fault. Compare fill time at a fixed switch-over pressure, not only peak. If the fast cavity fills 0.2 to 0.4 s sooner, the imbalance is hydraulic: unequal runner lengths, a gate 0.05 mm oversized, or shear heating in a runner corner lowering local viscosity. If fill times match but peaks diverge, the difference is packing or cooling: the hotter cavity stays open longer and packs more mass per bar.

Trace shapes separate the failure modes. A curve that rises late and peaks low points to a blocked gate or a vent fault, and its part shows a sink at the far end. A secondary bump during pack marks a jetting stream that closed late over the tip. The press position curve looks identical in every case.

The fix belongs to the tooling, and the sensor layer changes the feedback loop: a repair that used to be judged after 5,000 shots is judged on the next 20 traces. Medical and automotive customers expect the same per-cavity curves as IQ, OQ and PQ evidence, and the pressure archive is the hardest section of a validation package to fabricate and the easiest to trust.

Mounting: Flush Polish, Purge Rings and Wiring

A tip that is not mounted flush reads a plug of frozen plastic instead of the melt. Machine the bore to the tip outer diameter within 0.01 to 0.02 mm and seat the diaphragm level with the cavity surface. A proud tip protrudes into the flow, erodes, and leaks melt behind the sensor. A recessed tip freezes a dead meniscus that insulates the diaphragm, and the curve lags by the fraction of a second that matters at switch-over.

Polish the land around the tip to Ra under 0.1 micrometre and remove every lip. A step where the diaphragm meets the steel nucleates frozen skin and stresses the membrane edge at pressure peaks. On high-gloss or optical tools, match the surrounding polish so the tip leaves no witness ring in the part.

A spring-loaded purge ring solves the stuck-tip problem. The ring presses against the cavity surface and shears the part past the tip edge at ejection; without it, a shrinking part pulls the sensor from its bore or tears the diaphragm. Set the ring land 0.2 to 0.5 mm proud of the diaphragm so the part contacts the ring first.

Wiring is a mold-base detail that becomes a production liability when improvised. Run the bore from the tip seat through the plates to the mover side, exit with strain relief 150 to 300 mm from the mold surface, and clear the cable of ejector travel and slide paths. Use jacketed cable rated to at least 250 °C at the mold end, and keep the amplifier within its 2 m window on the mold. A tool that will carry sensors should say so on the drawing before the insert is finished, because a late bore cut from a drawing without the callout is the most common reason a retrofit stalls.

Closed-Loop Control: Pressure Profile vs Short-Shot Clamp

The short-shot clamp learns a peak or integrated-pressure window on good shots and alarms or rejects when a shot falls outside it. It changes nothing in the machine program and only gates the part. It catches viscosity drift, sink onset and pack failures, and it is the right first step on a validated process that needs a per-shot audit.

Pressure-profile control makes the switchover and the pack follow a cavity-pressure setpoint curve instead of time and position presets. The controller moves to pack when the tip reads the knee, holds the profile while the gate is open, and ends hold at freeze. Lot-to-lot viscosity drift no longer changes the part; it only changes how long the machine takes to follow the same pressure path. That is the difference between reacting to variation and removing it.

Hybrid triggering is the practical middle ground on most automotive cells. The switchover fires at whichever arrives first, the cavity-pressure preset or the position preset, so a fast lot saves cycle and a slow lot alarms. Pack then runs on the profile for its first 0.5 to 1.0 s while the gate is open, and on position or time after freeze.

Two cautions before closing the loop. A pressure-controlled pack removes the cushion's damping, so the clamp tonnage margin matters more, because the controller pushes until the cavity reads setpoint even as the tool starts to flex. And the loop can hide a degenerating tool: clogged vents and wearing gates get held to the curve while parts slowly worsen. Alarm on the correction effort the controller spends, not only on the part decision.

What the Sensor Layer Costs

The honest budget has four lines. A piezoelectric flush tip runs 500 to 1,500 USD depending on diameter, temperature rating and connector style. The charge amplifier adds roughly 300 to 800 USD per channel. Software ranges from about 2,000 USD per press for an alarm package to five figures for plant-wide SPC and archiving. Mold-side machining adds 100 to 300 USD per tip for the bore, the purge-ring land and the polish.

That machining line disappears when the bores are cut during build instead of retrofitted later. If the program needs no per-shot data yet, a drawing callout alone keeps a future retrofit to one wire-EDM pass rather than a teardown.

← swipe to scroll →

Table note: market-typical price bands as of 2026, quoted per channel or per tip. Validate against current vendor quotes for the exact tip family and software tier before budgeting a program.

A three-channel pilot lands near 4,000 to 8,000 USD installed, the number to compare against the cost it controls. Two percent scrap on a 16-cavity lid running 30,000 parts a day at 0.12 USD material throws away roughly 720 USD of resin daily before press time. Remove half of it in the first month and the layer pays for itself.

When Sensors Pay Off and When a Weight Check Is Enough

The decision is arithmetic, and it turns on three questions. Does cavity variation appear before part variation? Is the part valuable or regulated enough that a per-shot record has worth? Can a cheaper measurement catch the same fault? Weight checks, vision and dimensional sampling answer one question each, and the sensor answers all three every cycle.

Sensors earn their place in medical molding, where IQ, OQ and PQ documentation expects per-shot evidence and one unverified lot can stop a launch. They earn it in optics and cosmetic parts, where sink and the filling of a 0.2 mm feature are decided by peak pressure at the tip. No hourly sample catches the excursion between checks. They earn it in thin-wall packaging at 8 or more cavities, where cycles run 4 to 6 seconds. The costly scrap hides in the 300 shots between two weigh checks, which is what part weight variation control measures after the fact.

They are surplus where the fault is slow and the part is forgiving. A 4 mm wall housing in PP at ±0.2 mm reports viscosity drift through its weight, and a check every 30 minutes tracks that drift within 0.3 to 0.5 percent on a stable machine. When an hour of bad shots costs less than the amortized sensor set, the weight check is the control and the sensor is a ticket you did not need.

The middle path is temporary instrumentation. Magnetic-base tips and quick connectors let a qualified tool carry a setup through a lot-changeover study, and permanent channels are cut only if the data proves the fault mode real.

Retrofit Checklist for an Existing Mold

Adding a sensor layer to a tool that shipped without one is a mold repair project, and it fails on planning more often than on machining. Work the list before the tool reaches the shop.

• Fix the question first. Gate freeze, balance, short shots or stickiness each set a different tip, family and channel count, so a retrofit without the question buys the wrong sensor.
• Confirm steel access. The bore must reach the cavity face without crossing a water line, ejector bore or slide track within 15 to 20 mm. Mark the clearance circles on the drawings before quoting.
• Check wall distance. Leave 1 to 1.5 times the bore diameter to the nearest cavity surface. A 6 mm tip that violates it moves the plan to a gate-land differential or a runner tip instead.
• Machine the fit to the sensor, not the catalog. Ream or grind the bore to the actual tip outer diameter within ±0.01 to 0.02 mm, with seat depth setting the diaphragm flush.
• Plan the purge ring land. The 0.2 to 0.5 mm proud ring needs its own circular land, or the first retrofit tears its tip on shot 400.
• Route the wire before the trial. A drilled exit with strain relief 150 to 300 mm out, clear of ejector travel, saves a second mold teardown.
• Requalify, do not reuse. After any bore the tool is new: run a short trial, re-measure the part, and rebuild the pressure windows over 50 to 100 shots before the alarms go live.
• Update the tool file. Drawing, spares list and maintenance schedule gain the tip model, serial, calibration date and replacement interval. A sensor nobody tracks becomes a sensor nobody trusts.

The first production shots after a retrofit belong in the same review as a T1 sample set: traces archived, parts gauged, results recorded before the alarm limits are trusted to sort parts. That record slots directly into the first article inspection report for the program.

Cavity Pressure Sensor FAQ

Q1. How many sensors does a mold need?

Start with one or two. A gate-land tip plus a last-to-fill reference covers fill, pack, freeze and stick on a simple tool. Add one tip per cavity above 8 cavities, or on optics and medical work, at roughly 600 to 1,800 USD installed per added channel.

Q2. Can cavity pressure replace the machine's controls?

No. It supervises them. The machine still meters, injects and packs, and the sensor decides when each phase ends and whether the result passes. Controllers integrate the signal as an external setpoint or trigger input.

Q3. Why did my new tip read flat for the first ten shots?

The usual causes are a recessed diaphragm trapped behind frozen skin, or a tip in dead steel the flow front never touches. Check flushness with a feeler gauge at the bore edge and compare the flat trace against the fill sequence. A proud tip shows the opposite symptom: an eroded, noisy trace.

Q4. How often should sensors be calibrated?

Annual calibration against a deadweight tester is the common rule, with a verification whenever the trace shape changes without a process change. Strain-gauge units drift enough to move a ±5 percent alarm window off target within a quarter.

Q5. Does a cavity pressure sensor affect the part?

Only if mounted wrong. A flush tip with a purge ring leaves no witness mark; a proud tip leaves a cold slug dimple and erodes; a recessed tip traps a pip. Specify Ra under 0.1 micrometre at the land and inspect the first shots under low magnification.

The Bottom Line

A press curve records what the machine planned; a cavity pressure curve records what the part received. The gap is measurable: 0.5 to 1 percent repeatability at the peak against 5 to 10 percent scatter on clamp and velocity signals that never touch the melt. One tip converts switch-over, the peak window, gate freeze and ejection into per-shot facts, and a 10 percent cavity spread names an unbalanced tool the shift it appears.

The layer is not free and not always necessary. Budget 600 to 1,800 USD installed per tip plus the electronics, and weigh that against hourly weight checks on forgiving parts. Medical, optical, thin-wall at 8 or more cavities, and tier-1 automotive programs are where the sensor set earns its place.

DieStrike builds and retrofits injection molds to IATF 16949 with ±0.005 mm mold tolerance, ±0.002 mm on EDM and jig-grinding work, and 120+ machines behind 2 to 5 week tool lead times. Sensor bores, purge-ring lands and wiring paths are machined to drawing during build, so the instrumented retrofit never has to happen. Send the part data and the cavity count with your RFQ and we will return a tooling concept with the sensor layer priced in.