The diagnostic sequence “First Article Fails on a New PET Mould: Reading Wall Distribution” is applied when a multi-cavity injection mould for polyethylene terephthalate (PET) preforms produces a nonconforming first article. In a 24-cavity tool trialled on a 250-tonne electric injection moulding machine, the resin was an injection-grade PET with an intrinsic viscosity of 0.80 dL/g measured according to ISO 1628-5. The resin was dried to a residual moisture content below 50 ppm in a desiccant dryer with a dew point of −40°C. Melt temperature at the nozzle was 278°C, and the water-cooled cavity was held at 10°C. Wall thickness mapping with a Hall-effect thickness gauge having a resolution of 0.001 mm showed a mid-body circumferential variation of 0.14 mm and an end-cap eccentricity of 0.32 mm. The distribution was not stochastic; the polar trace followed a single sine wave, with one thick side opposite one thin side. That pattern is characteristic of core deflection during filling, not of a uniformly short shot or a gross hot-runner imbalance. Adjacent cavities at the same angular position agreed within 0.04 mm, but the thick–thin axis rotated across the cavity plate, indicating that the core shift was driven by local gate-pressure asymmetry rather than by a bent core pin common to all cavities.

The first-article failure was read in the wall distribution before any dimensional correction was made to the mould. For PET preforms, wall distribution is not a single thickness value; it is a map of the part’s polar thickness profile at defined axial levels. The failure signature is interpreted from the amplitude and phase of the thickness trace. A symmetric, flat distribution with low range is usually accepted, while a distribution with a single cosine peak indicates an off-centre core or an asymmetric gate. A distribution with a thick end cap and thin sidewall may indicate gate freeze during packing. The preform must be conditioned at 23°C and 50% RH for at least 24 h according to ISO 291 before measurement, because post-mould shrinkage and moisture uptake can shift the wall trace by up to 0.02 mm.

Polyethylene terephthalate is a semi-crystalline polyester, but preforms for injection stretch-blow moulding are normally quenched into an amorphous state. The glass transition temperature of dry PET is 78°C measured by ISO 11357-2. The cold crystallisation exotherm peaks at 140°C to 160°C, and the melting endotherm is typically 247°C to 255°C. For a preform moulded at 278°C and quenched to 10°C, crystallinity remains below 5%. Amorphous PET has a density near 1.34 g/cm³ per ASTM D792. This is relevant to wall distribution because the amorphous part is dimensionally stable only after conditioning; premature measurement can confuse thermal shrinkage with mould misalignment. The wall distribution reading is therefore performed after the part reaches conditioned dimensions, and the same zero point is used on the polar plot for all cavities.

What Wall Distribution Data Are Collected on a Failed First Article?

The wall distribution data set consists of three axial measurement planes—neck/shoulder, mid-body, and end cap—each sampled at 8 circumferential positions spaced 45°. The polar plot is referenced to a marked angular datum on the preform, not to the mould parting line. The failure signature is classified by three parameters: average wall thickness, circumferential thickness range, and eccentricity calculated as (t_max − t_min)/2. A first article failing for wall distribution typically exhibits a range above 0.10 mm or an eccentricity above 0.15 mm at the preform mid-body, although the acceptance window is product-specific.

The tolerances above are product-specific and must be derived from downstream stretch ratios and burst requirements; they are not universal acceptance limits. The table is a representative protocol for a 12.5 g preform with a 28 mm neck finish. The wall distribution map is read against the cavity axis rather than against a fixed datum on the preform. The coordinate measuring machine establishes the centreline of the neck and the end cap, and the Hall-effect gauge then measures the local wall thickness at each polar location. The difference between the maximum and minimum thickness is not enough; the angular location of the maximum indicates the direction of core displacement. A core shifted toward the parting line can produce a thick-thin pattern that rotates as the tool is indexed, while a core shifted toward the gate produces the same phase in every cavity. In the failed tool, the phase rotation across the cavity face indicated that the core deflection was not caused by a common bent pin but by a locally variable side-load.

On the failed tool, the mid-body trace showed a maximum thickness at one angular position and a minimum 180° opposite, with a repeatability of ±0.02 mm on two consecutive cycles. The absence of random variation in the peak location ruled out inconsistent transfer position. The wall distribution therefore shifted the investigation to the filling phase rather than the cooling phase. The ratio of flow length to wall thickness for the preform mid-body was 14:1 to 18:1, which is not an extreme thin-wall condition but is sufficient to make the core pin sensitive to gate-pressure imbalance. The 12.5 g preform had a nominal mid-body wall thickness of 2.10 mm and a flow length of approximately 95 mm from the gate to the end cap. At this thickness, the filling phase is pressure-limited rather than temperature-limited, and packing pressure dominates the final wall distribution.

The solidification sequence in a PET preform tool explains why wall distribution fails first and why process changes can mask a mould defect. At a melt temperature of 278°C and a cavity wall temperature of 10°C, the contact skin freezes within milliseconds. The core pin, with a length-to-diameter ratio above 6:1, remains elastic under the filling pressure but can be displaced by a local melt pressure difference of 4 MPa to 8 MPa across the annulus. The displacement is largest near the end cap, where the core is least supported. This produces the end-cap eccentricity observed in the first-article wall map. A Hall-effect thickness gauge cannot directly measure core deflection; however, the polar thickness trace can be converted to an eccentricity vector. The vector angle corresponds to the direction of core shift, and the vector magnitude corresponds to the deflection once the wall eccentricity is divided by two.

In PET, the shear rate at the gate can exceed 10,000 s−1 during the initial injection velocity spike, while the shear rate at the end cap is below 100 s−1 at the end of fill. PET is shear-thinning; its apparent viscosity decreases as the shear rate increases. A melt at 278°C with an IV of 0.80 dL/g can have an apparent viscosity above 50 Pa·s at 1,000 s−1. Small changes in injection velocity therefore produce larger changes in filling pressure near the gate and larger side-load on the core pin. Reading wall distribution provides a way to infer the pressure field that existed during filling, because the frozen wall distribution preserves the imbalance. The wall distribution reading is therefore a surrogate for core dynamic deflection, not a replacement for toolroom dial indicator checks. A dial indicator mounted on the core pin after the mould is split may show a static runout of only 0.01 mm; the dynamic deflection under melt pressure can be an order of magnitude higher.

Outer Skin Solidification, Gate Land Freeze, and Core Deflection in the Mould

Outer skin solidification and gate land freeze define the time window in which wall distribution can be corrected. In PET preform moulding, the gate land diameter is typically 0.8 mm to 1.2 mm for a valve-gated tool. With a water-controlled cavity at 10°C, a land of 0.8 mm can freeze before 0.5 s of hold pressure has been applied, depending on the nozzle tip temperature and the valve-pin closing delay. If the gate freezes early, the end cap receives insufficient packing and the wall trace thins at the dome. If the gate remains open too long, the core pin can be pushed against the cavity wall before solidification locks the core in place. The wall distribution on the failed first article showed mid-body eccentricity, not simply an underpacked dome, so the gate-freeze time was not the first correction target.

The pressure required to pack a PET preform is usually 30 MPa to 50 MPa hydraulic or 80 MPa to 120 MPa melt pressure, depending on the machine and screw diameter. The injection velocity during filling was initially maintained at 35 mm/s to 55 mm/s; this range can be too high for a long core pin because the core receives the full force of the melt stream from one gate. Data from the first trial showed that reducing screw velocity from 55 mm/s to 38 mm/s during the first 0.4 s of fill reduced the mid-body eccentricity from 0.32 mm to 0.21 mm. The remaining error was attributed to hot-runner thermal imbalance.

Cavity pressure transducers provide an independent check on the wall distribution reading. In the failed trial, a 1 mm quartz cavity-pressure sensor placed in the cavity wall gave a peak filling pressure of 62 MPa in cavity 3 and 48 MPa in cavity 14. The pressure difference of 14 MPa across the face was larger than the 6 MPa to 8 MPa side-load needed to deflect the core pin. The wall map confirmed the pressure transducer data; the cavity with the highest pressure produced a thick wall on the gate side and a thin wall opposite, while the cavity with the lower pressure showed a more centred core.

Cooling water balance across the cavity plate matters because PET solidification is governed by the heat transfer to the water circuit. A temperature differential of 2°C in the cavity cooling water can change the local skin thickness by a measurable amount, but it rarely creates a sinusoidal wall distribution unless the differential is asymmetrical around the core. In the failed tool, the cavity plate inlet temperature was 10°C and outlet temperature was 12°C on one side, while the other side showed 10°C to 11°C. This imbalance was not large enough to explain the observed eccentricity but was corrected along with the hot-runner settings.

Gate freeze time can be estimated from the gate land thickness and the thermal diffusivity of the PET melt, but the practical measurement is a pressure trace at the gate. When the cavity pressure behind the gate stops responding to additional hold pressure, the gate is frozen. On the failed tool, cavity 3 still responded to hold pressure for 1.2 s, while cavity 14 showed no response after 0.7 s. This difference of 0.5 s in gate-freeze time created unequal packing and contributed to the flat across-face wall variation. The valve-gate closing delay was set to 0.2 s after transfer to allow the gate area to pack before the pin was seated, but the delay had to be trimmed per nozzle to offset the manifold temperature spread.

When Cavity-to-Cavity Viscosity Gradients Make a Mould Look Defective

When cavity-to-cavity viscosity gradients exist, the wall distribution reveals flat but shifted average thickness values rather than a sinusoidal pattern. In the failed tool, the cavity-to-cavity average mid-body thickness varied from 1.98 mm to 2.31 mm, with the thinnest parts concentrated on one manifold branch. Manifold tip temperatures measured with a hot-runner thermocouple showed a spread of 6°C across the 24 nozzles. Because PET viscosity is strongly temperature-dependent near the processing temperature, a 6°C tip-temperature difference changes the local fill speed and the packing pressure inside that cavity. This shifts the average wall thickness independent of core deflection. A first-article wall distribution should therefore be read twice: once within a single cavity as a polar trace, and once across the cavity face as an average-thickness control chart. The within-cavity trace identifies core movement; the across-face chart identifies hot-runner balance.

Resin condition also changes the wall distribution. If PET is processed without drying to below 50 ppm moisture, hydrolytic degradation lowers the intrinsic viscosity from 0.80 dL/g toward 0.70 dL/g under melt residence times longer than 3 min. The lower-IV material fills faster and packs differently, producing thinner walls in the first-filled cavities and thicker walls in the later-filled cavities. Acetaldehyde formation also increases above 280°C when residence time exceeds 5 min, but acetaldehyde does not directly affect wall distribution; it is a compliance marker under FDA 21 CFR 177.1630 and Regulation (EU) 10/2011. Intrinsic viscosity is measured in a 60/40 phenol/1,2-dichlorobenzene mixture at 25°C per ISO 1628-5. A change of 0.01 dL/g is analytically significant; a change below that is often within interlaboratory repeatability. The first-article wall map may shift before the IV change is detected if the hot runner amplifies a small viscosity difference into a pressure imbalance.

The correction expected from each action is a range because wall distribution response depends on the preform geometry, core pin rigidity, and hot-runner layout. Published data for the specific interaction between IV reduction and preform wall distribution in a 24-cavity valve-gated tool is limited; the relationships described are inferred from cavity pressure and part geometry rather than extracted from a single standard.

The consequence of wall distribution failure in the preform is not only dimensional; downstream stretch ratios amplify local thickness differences. A preform mid-body eccentricity of 0.09 mm can become a bottle sidewall variation above 0.03 mm after axial and hoop stretching, depending on the stretch ratios. The axial stretch ratio for a 28 mm finish PET bottle is typically 2.5 to 3.0, and the hoop stretch ratio is 3.0 to 4.0. The local bottle wall thickness can be approximated by dividing the preform wall thickness by the product of the axial and hoop stretch ratios; therefore, preform wall distribution is a direct input to burst and top-load performance.

Correction on the failed tool proceeded from the hot runner toward the core. The manifold tip-temperature spread was reduced from 6°C to 2°C by replacing two nozzle heaters and recalibrating the thermocouple inputs. The injection velocity profile was set to 38 mm/s for the first 0.4 s and then 50 mm/s until transfer at 98% part volume. Hold pressure was set to 45 MPa melt pressure for 2.0 s, and the cavity water supply was balanced to 10°C ± 1°C.

After the hot-runner tip spread was reduced to 2°C and the injection velocity profile was changed, the first-article wall map showed a mid-body range of 0.06 mm and an end-cap eccentricity of 0.09 mm. The core pin deflection was then checked against a toolroom cylindrical gauge and found to be within 0.02 mm. The cavity-pressure sensors recorded a spread of 5 MPa across the face instead of 14 MPa. Because PET wall distribution responds to both melt pressure and core rigidity, the measured improvement was not assigned to a single variable; the data were recorded as a combined effect of hot-runner balance and fill-speed reduction.