Cpk Over Certificates: Why Process Capability Is the Real Test of Aerospace-Grade IN718
By Harris, Technical Engineer, FUSHUN METAL
Here is an uncomfortable truth about superalloy procurement: a mill test certificate proves almost nothing about the next heat you will receive. It tells you that one batch, tested once, passed a specification. It says nothing about whether the supplier’s process can repeat that result a hundred times without drifting toward the limit. In aerospace, where a turbine disk failure is not a warranty claim but a catastrophe, that distinction is everything. My position is simple: for engine-grade IN718 (a precipitation-hardened nickel-iron superalloy, also known as Inconel 718 or UNS N07718), buyers should stop asking “did it pass?” and start asking “what is your Cpk?”
A 2025 study published in the Chinese journal Special Steel Technology makes this case with unusual clarity. It analyzed three production heats of Φ200 mm forged IN718 bar qualified to AMS 5662N for aircraft engine use, and instead of merely reporting pass/fail results, it computed process capability indices for chemistry and tensile properties. The numbers deserve attention from anyone who buys or specifies this alloy.
What Cpk Actually Measures
Cpk quantifies how comfortably a stable process sits inside its specification limits, expressed in units of process spread. It takes the distance from the process mean to the nearest specification limit and divides it by three standard deviations. A Cpk of 1.00 means the process just barely fits; roughly 2,700 parts per million will fall outside limits. A Cpk of 1.33 — the widely accepted industrial minimum — corresponds to about 63 defective parts per million. At 1.67, the defect expectation drops below one per million.
Most industry commentary treats 1.33 as a finish line. I disagree. For flight-critical rotating hardware, 1.33 should be the entry ticket, not the target. The difference between a supplier running at Cpk 1.35 and one running at 3.0 is the difference between a process that will occasionally surprise you and one that statistically cannot.
The study’s chemistry results show what a genuinely capable process looks like. Every element that defines IN718’s metallurgy sits near the middle of its specification window, and even the tightest-controlled element — niobium, which forms the strengthening gamma double-prime phase — clears the sub-one-per-million threshold:
| Element | Spec lower | Spec upper | Measured mean | Cpk |
|---|---|---|---|---|
| C | – | 0.080 | 0.025 | 1.52 |
| Cr | 17.00 | 21.00 | 18.07 | 8.64 |
| Mo | 2.80 | 3.30 | 2.97 | 4.38 |
| Nb | 4.75 | 5.50 | 5.38 | 1.73 |
| Al | 0.20 | 0.80 | 0.53 | 3.33 |
| Ti | 0.65 | 1.15 | 0.96 | 4.34 |
| Ni | 50.00 | 55.00 | 53.94 | 1.71 |
The lowest chemistry Cpk across the board is 1.71 — every element beyond the 1.67 world-class threshold, and chromium at 8.64 is a process so centered that a nonconforming heat is a statistical fiction.
Tensile Capability: Where the Argument Gets Sharp
Chemistry is the easy part; modern vacuum melting controls composition well. The harder question is mechanical consistency, because tensile properties integrate everything — melting, forging reduction, grain size, heat treatment response. Room-temperature tensile results across three heats, sampled at head and tail of each bar, clustered between 1,430 and 1,458 MPa ultimate strength and 1,189 to 1,246 MPa yield strength. The capability indices, computed per the WP4902 consistency procedure in Minitab with Box-Cox and Log-Normal transforms where distributions required them:
| Property | Spec minimum | Observed mean | Cpk | Requirement |
|---|---|---|---|---|
| RT ultimate tensile strength | 1,241 MPa | 1,406 MPa | 4.80 | > 1.33 |
| RT yield strength | 1,000 MPa | 1,203 MPa | 3.51 | > 1.33 |
| RT elongation | 6 % | 16.98 % | 2.75 | > 1.33 |
| 650 °C ultimate tensile strength | 965 MPa | 1,120 MPa | 4.39 | > 1.33 |
| 650 °C yield strength | 841 MPa | 995.6 MPa | 3.16 | > 1.33 |
| 650 °C elongation | 6 % | 15.2 % | 1.73 | > 1.33 |
The 650 °C results matter most, because 650 °C is IN718’s practical service ceiling — above it, the strengthening phase coarsens into the ineffective delta phase and strength collapses. A process that holds Cpk above 3 at the alloy’s most demanding operating temperature is not passing a test; it is demonstrating margin that a certificate can never show.
Ppk: Capability Without the Flattering Assumptions
Cpk assumes the process is in statistical control and uses within-subgroup variation. The skeptic’s counter is fair: what if the process drifts between heats? That is what Ppk answers — it uses total variation, within and between batches, and does not require the process to be in control. It is the harsher judge. The study computed Ppk on 24 samples per property:
| Property | Lower limit | Mean | Ppk | Requirement |
|---|---|---|---|---|
| RT ultimate tensile strength | 1,276 MPa | 1,443.57 MPa | 6.33 | > 1 |
| RT yield strength | 1,034 MPa | 1,209.5 MPa | 2.35 | > 1 |
| 650 °C ultimate tensile strength | 1,000 MPa | 1,159.67 MPa | 3.84 | > 1 |
| 650 °C yield strength | 862 MPa | 1,020.8 MPa | 1.75 | > 1 |
Note that the Ppk calculation applied tighter lower limits than the base specification — 1,276 MPa instead of 1,241 for room-temperature ultimate strength — and still returned 6.33. When capability survives both a harsher statistic and a harsher limit, the conclusion is no longer arguable.
Why Certificates Alone Mislead
Consider two suppliers, both certifying yield strength of 1,050 MPa against a 1,000 MPa minimum. Supplier A’s process mean is 1,055 MPa with tight scatter; supplier B’s mean is 1,200 MPa with the same scatter. Both certificates look identical. Yet supplier A is one bad heat away from a rejection — or worse, a marginal heat that passes the coupon test but carries local weak zones the coupon never sampled. Certificates report a point; capability indices report a distribution. Only the distribution predicts the future.
This is precisely why engine OEMs such as GE, Rolls-Royce and Safran layer proprietary specifications on top of AMS 5662, demanding statistical evidence of process control rather than isolated test results. There is also a head-to-tail argument buried in the study’s raw data: samples came from both ends of each bar, and the spread was negligible — batch 1 room-temperature ultimate strength ran 1,430–1,450 MPa at the head and 1,430–1,450 MPa at the tail. Positional uniformity of that order cannot be inspected in; it can only come from a controlled process.
What Buyers Should Do Differently
First, ask suppliers for capability data, not just certificates. A mill that runs statistical process control will have Cpk figures for key chemistry and tensile properties and should be willing to share them under NDA. A mill that cannot produce them is telling you something important.
Second, set the bar at 1.67, not 1.33, for flight-critical properties. The published data above proves that well-run IN718 production lines achieve Cpk between 1.7 and 8.6 routinely. Accepting 1.33 as sufficient means accepting a process with two orders of magnitude more expected escapes than the state of the art delivers.
Third, treat capability data as a comparison tool between suppliers. Price per kilogram is visible; the cost of a marginal heat discovered after machining is not. In our experience at FUSHUN METAL, the suppliers who publish or share capability indices are consistently the ones whose material machines predictably, heat-treats predictably, and never generates the Friday-afternoon quality escape. That correlation is not a coincidence. It is what statistical control looks like from the customer’s side of the purchase order.
