Split DoE

[#1]

Hi,
I designed DoE by RSM. I wonder if I can split the design and first step to try the main effects and then 2nd step to try the rest experiments for the 2nd order effects.
Please advice

[#2]

If you do this, you are running parts of the design in 2 blocks and therefore might (likely) need to account for the block effect. 

I don't know anything about your situation, like the number of factors you are looking at, but typically if you are doing RSM, you already have data to support a first order model AND you have a good amount of information about noise.  If not, perhaps you can do sequential experimentation and start with fractional factorials and block designs.

[#3]

RSM is an old method and still valid. The Central Composite Design method, in particular, had an advantage of running the two-level treatments and some of the center points first, and then deciding if continuing with the axial points was warranted.

On the other hand, modern methods offer another way. You can create a small custom screening design and then augment the design to address deficiencies such as low power or model bias. You could also use the economical Definitive Screening Design method if it is a screening scenario (e.g., determine a small number of important factors out of a large number of candidate factors.) These designs support estimating linear, non-linear, and interaction effects when the screening principles hold.

You have choices.