The statement that a process can be controlled “within an acceptable range” requires some definition. A recent study by Hitchiner and McSpadden (2005) investigated the process variability of various vitrified cubic boron nitride (CBN) processes as part of a larger program to develop improved wheel technology. They showed that under “ideal” conditions repeatability of wheel life within ±15% or better could be achieved. However, variability associated with just wheel grade from one wheel to another (±1% porosity), all within the standard limits of a commercial specification, made the process less repeatable and increased the variability to ±25%. In the field, for example, in a high-production internal-grinding operation with 20 machines, the average monthly wheel life was tightly maintained within ±5%. However, these average values obscured an actual individual wheel life variability of ±100%! Of these, wheels with very low or zero life were associated with setup problems while the large variability at the high end of wheel life was associated with machine-to-machine variables such as coolant pressure, spindle condition, or gauging errors. A process apparently in control based on monthly usage numbers was actually quite the opposite (Figure 2.1 and Figure 2.2).
Wheel makers and machine tool builders are usually in the best position to make predictions as to wheel performance. Predictions are based on either laboratory tests or past experience on comparable applications. Laboratory tests tend to reproduce ideal conditions but can make little
allowance for a deficiency in fixturing or coolant, etc. In fact, the author witnessed a situation where the laboratory results and the actual field wheel life differed by a factor of 40. The loss of wheel life in the field was caused by vibration from poor part clamping and wheel bond erosion from excessively high coolant pressure. Laboratory data were able to inform the end user that there was a major problem and provide evidence to search for the solution.