Optimal Top logy for Distribu d Fault Detection of Large-scale Systems