[Ted Karr]

I agree with Robert’s reply and general questions. As a general model,  I like the approach of going to a stadium for a ball game.

1. What stadium am I in? Wrigley Field or Candlestick. A screening DOE as a guide for factors and levels.

2. What Section are the seats located? The screening gives some insight but know you are judging where are we and where do we need to go? So, DOE #2 is a refining step.

3. What is my row and seat number? How exact must we be? Understanding main factors, or interested in possible interactions, or is a prediction model required.

4. We ok as is or do we want Club seating with AC and food delivery service? This would a very refined design with just a few critical factors.

Most DOE teams are very excited to see the results from step 2. This either confirms assumptions or dispels them or gives an insight not previously considered. Most techies want to go to step 3 for no other reason than curiosity.

Screen, sharpen, hone, sticking it!

A great tool that often yields surprises.

Ted

 

 

 

0

[Ted Karr]

Bob,

Great reply. I would also ask for a definition of productivity.  Max output or max output at minimal cost or throughput times?

Yes, you absolutely must know if the A,B, and C items are 1 per batch or is there a combo of raw materials? Additionally, do the catalysts function with all the raw materials for their desired effect? If not, then there will be constraints on the DOE factor levels.

To roughly apply a Shainin method you may want to run all factors at their lowest cost level, at the mid level cost, and at the high cost end. After all, the typical goal of DOE is to develop a product that meets specifications at the lowest cost. Those 1st 3 simple experiments could put you in the ballpark for future experiments. For instance, if the middle cost catalyst works the best, you can now focus on using the 2 catalysts in the mid cost range. The same thinking can be applied to your raw materials. As a result you can cut all of your factors in half with this approach.

These ideas are based on supplied information. They may be invalid depending on supplied future details.

• Ted

0

[Ted Karr]

Jonathan,

Take measurements of multiple parts from each cavity and make a box plot of EACH cavity. This is a multivariate chart and will clearly display the part variation within and between cavities. If a difference appears likely then you can run a hypothesis test on the mean values of each cavity to determine if they truly are different. Use the great Minitab assistant screens to guide you and work the examples using their preloaded data sets as referenced in the help screens.

Ted

0

[Ted Karr]

Rob,

Two sources…1. IHI (Institute for Healthcare Improvement) and 2. Am Soc for Quality which has a $69 on line class with group discounts. Not too sure you would need 3 to 5 days to cover this material even with case studies. I would think 2 days would be sufficient.

Tedd

0

[Author]

Posts