Experimental Design

[memilanuk]

John Stranahan has a good book "Precision Rifle Rebarreling and Case Preperation on a Home Shop Lathe" He has a chapter dedicated to "load development, factorial design, accuracy loads. He charts bullet jump versus powder charge to plot best accuracy. Once you understand the principals, you can use it for other reloading components and variables. I use it to develop my loads. I think it saves time and the number of rounds fired to find the best load.

Actually, I do have that book squirrelled away here. I'd forgotten he'd touched on basic 2k factorial testing (mostly 2 x 2 grids). I had been curious about the simplex optimization techniques he mentioned after the factorial section. In the example Stranahan provided, it sure seemed slick. At some point I think I asked someone (who did DoE for a living) and they had a less than glowing opinion of simplex optimization in general, so I kind of let it drop. Still it is an interesting concept; maybe I'll have to give it go at some point.

Monte

 

[Warren Dean]

haven't been to England for the IQ approved intelligence tests, but maybe one day.. is Chicago, closer an around Halloween they feature a costume party for which many members create pun-based costumes an the annual Mensa Mind Games competition.. Oh Yea !

whaddayathink

Sheldon C

I think your grammar is atrocious.

 

[RegionRat]

Have you considered a slight variation such as a Taguchi method? You can miss events and confuse yourself if you don't spend some time studying the importance of the levels and interactions.

Nothing wrong with factorial designs. I have just seen runs where important factors were dismissed or passed over too easily when two levels should have either been broken into more steps or set farther up or down the scale.

 

[memilanuk]

I'm open to a lot of things... Plackett-Burman, Box-Behnken, D-optimal designs, etc. so long as they are appropriate.

I've heard of Taguchi; I understand the general goal - to design tests that are considered 'robust' in the face of real-world 'noise' that can't be controlled to any degree. I've seen mention in a few texts that the reception of the methods behind that idea has varied considerably. Again, I've been warned away from these in the past. Doesn't mean I'm not interested in seeing how they work for myself, though.

As I mentioned to someone else off-thread... it took me a long while to really 'get' the idea of fractional factorials and how aliasing works - and I'm still operating on theoretical belief, not practical experience, at that.

 

[RegionRat]

Then, I would say it is time to start your plans. Like I said, nothing wrong with factorial designs. The true hard part is learning where to set levels.

The slight difference between factorial and Taguchi isn't worth worrying. With load development, things can get violent and expensive if you set levels too far.

What type of goal do you have? Are you just trying to develop a specific new rig or load?

 

[memilanuk]

At first, just tuning either existing loads or fine tuning new ones. Mostly just curious about possible interactions between some of the factors we vary - say seating depth and powder charge, that we traditionally vary independently - which may mask the interaction effect.

Later... well, I've been collecting several different kinds of .223 Rem brass, some (more) 50-55gn bullets, appropriate powders, primers, etc. Now I just need a slow-twist barrel to with them. Thinking of doing a full-on screening test just to see where it leads.