Contents
- Index
VarTran's Capabilities
VarTran® is a powerful new tool for obtaining robust designs. It provides both a simpler and more powerful alternative to Taguchi Methods. It has the following advantages over using the inner/outer arrays and signal to noise ratios proposed by Taguchi:
(1) If an equation is available, VarTran can use the equation directly. In this case, VarTran frequently gives exact solutions. Using the equation to calculate values for Taguchi inner/outer arrays results in loss of important information including the effects of interactions.
(2) When no equation is available, VarTran can use the equations generated from response surface studies. These equations can include interactions. Again Taguchi inner/outer arrays can miss the effects of these interactions. Further, response surface studies are very efficient in terms of the number of experimental trials run. Six input variables can be studied in 32 trials. For inner/outer arrays, the total number of trials is the number of trials in the inner array times the number of trials in the outer array. Studying these same six variables requires a minimum of 72 trials (12 x 6).
(3) Both approaches require estimates of the input's variation. These estimates are used to design the Taguchi inner/outer arrays. This means that exploring alternate tolerances including tightening of tolerances requires additional data. VarTran does not use the estimates of the input's variation until the analysis is performed. This allows alternate tolerances to be freely explored without requiring further data.
(4) The Taguchi inner/outer arrays approach is akin to statistical tolerancing. Statistical tolerances assume that the inputs remain centered, an assumption that is frequently violated in practice. VarTran allows any method of tolerancing to be used.
(5) The accuracy and precision of the resulting estimates of the variation using VarTran are far superior to those obtained using Taguchi methods.
While VarTran provides a superior approach to achieving robust designs, this approach is entirely consistent with Taguchi's philosophy of starting with low cost tolerances, considering variation when setting targets to achieve robust designs, and tightening tolerances only as a last resort.
VarTran also performs tolerancing. It performs both worst-case tolerancing and statistical tolerancing. More importantly it implements a powerful new approach to tolerancing called process tolerancing. Process tolerancing provides a unified approach to tolerancing encompassing both of the previous approaches. It allows worst-case tolerances to be specified for some inputs while statistical tolerances are specified for others. Combinations of the two types of tolerances can be specified for still other inputs. No longer is one forced to pick between statistical or worst-case tolerancing. The result is accurate tolerances the correctly describe the performance of the product or process.
Finally, VarTran is a powerful math package capable of evaluating and plotting equations, and performing optimizations. It includes a powerful new approach to optimization called interval analysis. The interval analysis routines implemented in VarTran are guaranteed to converge to the global optimum regardless of the number and location of local optimums. Further, the optimum is found even if the function is not defined over much of the region of search. So even if you never use VarTran's capability to predict variation, you can still make use of its superior optimization routine and plotting capabilities. Following a designed experiment, VarTran can be used to further explore the fitted equations and to find optimal targets.
Traditional math packages specify values for the inputs. VarTran differs in that it allows you to specify behaviors instead. The result is that it calculates behaviors for the outputs instead of just values. This is the difference between how the product or process is suppose to perform versus how it will actually behave.
For further information, see What VarTran Does and Getting Started.