03 / Process validation · Film-to-film bonding
All 33 bags held pressure. Could I sign off five sizes?
For a common bonding process, I needed to know which settings made a joint reliable. Eighteen 1 L bags and fifteen 50 L bags held pressure. The forms did not say which bond settings produced those bags.
The missing field changed the release decision.
The smallest and largest bags bracketed the family’s seal lengths. If both extremes passed at the proposed settings, that would support using one process across sizes. The pressure tests showed the finished bags held up. They did not connect those outcomes to a particular machine recipe. I kept the common process as a proposal and put specimen traceability ahead of another broad experiment.
The temperature data changed my own explanation, too. One measured cycle spanned 16.0 °F between channel peaks. In a later ten-run screening set, the hot location moved. I withdrew the claim that a fixed hot spot set the usable window. A location on the fixture, its measurement history, and the outcome on the same specimen have to agree.
Measured / Rev F, Figure 17
One cycle. A 16 °F spread.
The coolest channel peaked at 228.9 °F. The hottest reached 244.9 °F. These are peaks from four channels during one cycle.
Why I revised the interpretation in Rev D
My early assessment favored a narrow corridor and treated the apparent location differences as evidence about the tool. Sensor placement and repeat runs weakened that inference. I kept the measurements and withdrew the claim that this spread alone established the width of the process window.
Physical model / transfer decision
What would let me release five sizes?
Heat has to cross the tool face, the contact surface and the film. Local pressure and contact time affect that path. A final pressure hold answers a different question about the finished assembly.
Keep the specimen attached to its evidence.
The 18 small and 15 large bags passed a stepped hold to approximately 6.5 psi. The thermal runs and the functional results did not identify a matching specimen in every proposed size and condition. I held the family-wide recommendation open and targeted the missing links before spending on repeat tests. Read the pressure-hold and revision tables.
Read the original coverage map
Why calorimetry alone could not settle the question
Calorimetry identifies material transitions. The joint also depends on time at the interface, contact and geometry. My release decision needs a linked functional result at each limiting condition.
Separate vendor tooling acceptance
The defect kept choosing the same edge.
The first explanation was post-bond cooling. A defect that recurred on one side made me ask for the heater plate’s first-article dimensional inspection. It identified an uneven face.
I required the vendor to retool before qualification. A shim might improve one setup while leaving every later build dependent on the adjustment.
Project records · Rev D process-window assessment, Rev F face-to-face DOE and separate tooling-acceptance account.