I have never worked with array-data before and I have som basic question on
my setup.
I have normalized data in a time series with two replicates.

*my samples: *

colnames(hela.bc)
[1] "I.0h"   "I.1h"   "I.2h"   "I.3h"   "I.6h"   "I.9h"   "I.12h"
"I.15h"  "I.18h"  "I.21h"  "I.24h"  "II.0h"
[13] "II.1h"  "II.2h"  "II.3h"  "II.6h"  "II.9h"  "II.12h" "II.15h"
"II.18h" "II.21h" "II.24h"

I would like to compare "I.0h" and "II.0h" against all the other time
points. (I=replicate 1, II=replicate 2)

I set up the following, but I wonder if its correct:
 

design <- c(2,1,1,1,1,1,1,1,1,1,1,2,1,1,1,1,1,1,1,1,1,1)
fit <-lmFit(hela.bc, design)
fit2 <- eBayes(fit)
topTableF(fit2, adjust="BH")*

No, that is not correct. Have you read the limma User's Guide? If you are planning to analyze data, you need to educate yourself (or better yet, find a local statistician to help; this isn't a trivial task to take on lightly).

Anyway you want to do something like

times <- factor(paste0("Hr", rep(c(0,1,2,3,6,9,12,15,18,21,24), 2)))
design <- model.matrix(~times)

and proceed from there. The parameterization for this model will be a comparison of each time point to the baseline at time zero.