Aerospace Composite Layup: Predictive SPC for Less Scrap

In aerospace composite layup, scrap is not a quality event — it is a process failure that occurs over time. Fibre orientation does not snap out of tolerance in a single pass. Gap width does not suddenly triple between plies. These conditions accumulate gradually, often over 15 to 30 consecutive tow passes, while static SPC charts show the process in control because the UCL was set six months ago against a different process distribution. By the time a point breaches the static limit, 8 to 12 panels may have already been produced with the same underlying drift — and every one of those panels is at elevated risk of post-cure rejection.

Predictive SPC closes this detection gap by replacing static limits with live adaptive boundaries that update with every new measurement. The ML model tracks not just whether a point is within the static UCL, but whether the process distribution itself is shifting relative to its own recent history. Western Electric rules are evaluated automatically on every data point — detecting runs, trends, and patterns that a single-rule ±3-sigma system cannot see. The result is detection at the point of drift onset, not at the point of limit breach. The difference for scrap reduction is a documented 30-50% reduction across mature predictive SPC deployments.

Traditional SPC is the regulatory baseline — no AS9100-registered operation passes an audit without documented control charts. But compliance and prevention are not the same thing. The table below shows the operational difference between running static SPC for compliance and running predictive SPC for scrap prevention.

Walter Shewhart's original control chart concept used a single rule: a point beyond the ±3-sigma control limit signals an out-of-control condition. The Western Electric rules, published in 1956, expanded detection to eight specific patterns — runs, trends, mixtures, and stratification that the single-rule system misses entirely. These eight rules are the statistical standard for aerospace quality systems, but they are almost never applied in practice because manually monitoring all eight rules across multiple parameters is mathematically impossible at production speed. Predictive SPC automates every rule on every data point, surfacing patterns that static SPC leaves buried.

Our static SPC system was monitoring exactly one rule: point beyond ±3 sigma. It never fired on the drift that caused our two post-cure scrap events last quarter. When we loaded the same data into a predictive SPC system that evaluates all eight Western Electric rules simultaneously, every drift event that preceded those scrap panels was flagged by rules 2, 3, and 5 at the time it was happening — 15 to 20 passes before the single-point breach would have occurred. The static limits were not wrong. They were just incomplete. We were running a quality system that only detects emergencies while ignoring the trends that cause them.

Process capability index (Cpk) is the most direct statistical measure of scrap risk. A process running at Cpk 0.9 generates 2,700 defects per million opportunities — corresponding to a first-pass yield around 88-91% typical of static SPC operations. Moving from Cpk 0.9 to Cpk 1.33 reduces the defect rate from 2,700 ppm to 64 ppm — a 42x improvement. At Cpk 1.67, the defect rate falls to 0.6 ppm. Predictive SPC delivers this capability ladder as a live metric updated with every measurement, not as a quarterly report. The supervisor sees the current Cpk for every critical characteristic on a single dashboard and knows exactly how much scrap risk the current process state carries.

Predictive SPC reduces scrap through three distinct mechanisms that operate at different levels of the production process. Each mechanism addresses a specific failure mode that static SPC cannot handle.

Predictive SPC deployment follows a structured pathway that begins with connecting to existing measurement data and extends to full adaptive control limit management. The system is designed to run alongside existing quality workflows without requiring data migration or process re-qualification.

The financial impact of scrap reduction through predictive SPC is not theoretical. The figures below are based on published aerospace manufacturing benchmarks and iFactory deployment data for a composite aerostructure programme running 200 panels per month at a programme value of $12,000 per panel.

Scrap in aerospace composite layup is not a quality event that happens at the autoclave. It is a process condition that develops across passes and plies, measurable in the data from the first tow pass that deviates from the centreline. Predictive SPC transforms scrap from a lagging indicator that arrives with the post-cure inspection report into a leading signal that supervisors can act on during production.

The difference between static and predictive SPC is not the statistics. The math Shewhart developed in 1931 is still valid. The difference is speed, automation, and intelligence. Static SPC requires a human to read a chart, spot a trend, and decide to act. Predictive SPC evaluates every rule on every measurement, surfaces the drift pattern at the earliest possible point, and suggests a corrective action before the characteristic reaches a value that triggers the static limit. The operations that deploy predictive SPC report 30-50% scrap reduction within 12 months — not because they changed their process parameters, but because they started seeing drift early enough to correct it.

iFactory's predictive SPC platform is purpose-built for AFP composite layup operations — integrating with existing measurement systems to deliver automated Western Electric rule evaluation, self-tuning control limits, live rolling Cpk, and pre-breach drift alerts that let supervisors prevent scrap rather than report it. Book a Demo to see predictive SPC running on a composite layup use case matched to your process profile, or Talk to an Expert to discuss scrap reduction targets for your specific programme.