I agree with all the references provided by @Victor_G . But if your focus is on the math, there is a mathematical basis for both creating designs (classic, optimal, mixture, along with accompanying techniques like blocking, indeed all DOE family types) AND the analysis. Because so much of the math depends on the exact modeling method. The math behind ordinary least squares is very different from say nominal logistic regression. Or what about using an associated modeling method like stepwise regression. Different math again. So be prepared.
And never ever go to the modeling without first exploring your experimental results graphically. So I'd also suggest learning all these techniques as well. This is quite frankly where JMP shines...sure the math is there...but by design JMP tries to keep the math in the background and let's you explore visually. Trust me...having worked at SAS with the developers and coders of JMP...they've forgotten more about statistics than I'll ever know...they get the 'math' right.
Math aside, if you are new to DOE in general I'd also try and seek out a mentor that has a deeper and broader understanding of DOE. It's one thing to read a book and do online exercises...it's a whole other animal actually doing it so find someone to hold your hand along the way.
I hope this helps?