Dennis Mok, Supplier Engineer, Applied Materials
Evaluating supplier equipment matching and modeling supplier process are key functions of supplier engineering. This presentation demonstrates how to standardize supplier equipment matching and showcases supplier process modeling.
In the example given, a supplier wanted to validate that a new bead blaster matches the existing bead blaster based on the resulting roughness when using the same aluminum material, operator, and process parameters. First, supplier data were verified as normal using normal-quantile plot and normality goodness-of-fit Test. To validate the matching requirements, two JMP platforms (Matched Pairs and One-Way ANOVA) were used and compared to evaluate roughness differences between the tools. After rejecting the null hypothesis, the equivalence test and the forest plot in JMP 17 were produced to make a practical decision of the difference. To further study the sources of mean differences, multiple comparisons were used to clarify the differences between parts and test locations.
Supplier also wanted to study the relationship between predictor (bead blast pressure) and response (roughness of ceramic materials) to find the best fitting model and optimal pressure setting. Fit Y by X bivariate platform was conducted and compared fitting algorithms (least square, Cauchy, polynomial) to address the lack of fit and non-random residual patterns. Quadratic fit was selected based on the R-square adjusted, lack-of-fit p-value and max R-square to identify the potential improvement within the constrained pressure range. Profiler option in the Fit Y by X platform in JMP 17 was utilized to maximize desirability to find the optimal pressure setting.
