Jason Robbins, Process Engineer, US Synthetic
Developing a nondestructive test (NDT) gauge can be extremely difficult in any industry. Using the Discriminant Analysis Platform in JMP made this challenging task possible. Nondestructive tests are important to us because our current destructive tests don’t always describe our population of products as a whole very well. Failing to deliver the correct product to our customers creates a domino effect of wasted time, money and dissatisfied customers. In this presentation I will show how we: 1) Were able to discriminate the results from several NDT tests. 2) Chose between the linear, quadratic and regularized models. 3) Used the Stepwise Variable Selection to reduce the number of covariates and simplify the model. 4) Proposed, validated and fitted models using Holdout Validation. 5) Saved formulas from the fitted model for classification of future observations. 6) Implemented the NDT to manufacturing to be used by the operators. Without the Discriminant Analysis platform in JMP we would not have this NDT. Discriminant analysis is important for us and will be for others because it was the key tool that allowed us to create this NDT gauge and solve our problem.
