Optimized Lapping Process on New Product Launch
For years my employer ( a major valve manufacturer) used a method of refining valve seats, which they referred to as “croaking”. On established products “croaking” worked well. On new product launches, however, ramp up to full capability was long and painful because tuning the process was handled OFAT (one factor at a time). I was concerned with the long ramp ups because I was launching a new product that relied on “croaking” to meet leakage requirements.
My concerns were well founded, an early capability run showed low process yield and lots of scrap. Discussions with the manufacturing group did not ease my mind, there seemed to be as many theories on the problem as there were people in the room.
I proposed that we use a designed experiment (DoE) approach to identify and optimize the key process control variables. One person in the group agreed.
Our test plan was to run a preliminary DoE to determine the best mean settings and, if needed, run a second DoE using log normalized variance to address process spread. For the first DoE, four control variables and one noise variable were selected. They were:
Key Process Input Variable
|Material/Heat treat||11L17 ,S-83||86L20, S-228|
|Croak Speed||90 rpm||180 rpm|
|Thrust pressure||15 psi||25 psi|
|Hours on Croak Ball (noise factor)||0 hrs||80 hrs|
The preliminary DoE showed that the edge chamfer was, by far, the most significant primary factor and that it also interacted with the material type. The results from the first DoE were good enough that we concluded that the second experiment wasn’t needed.
We changed the material to 80L20 and added the 0.030” chamfer. A 50 piece validation run yielded a Cpk of 1.72, better than our quality objective.