No-confounding designs of 20 and 24 runs for screening experiments and a design selection methodology