In　this　paper,　an　automated　approach　for　the　radial　inflow　turbine　preliminary　design　and　off-design　prediction　is　developed.　The　system　is　based　on　well-established　loss　models　and　the　genetic　algorithm.　The　total-to-static　efficiency　is　selected　as　objective　function,　and　the　best　combination　of　the　flow　coefficient,　loading　coefficient　and　the　rotational　speed　can　be　found　within　a　predefined　design　space.　The　design　case　published　by　Jones　in　1996　is　selected　for　the　validation　work.　The　results　prove　the　correctness　and　effectiveness　of　this　automated　preliminary　design　approach.　The　CFD　results　are　higher　than　the　experiment　data　because　of　no　consideration　of　windage　and　mechanical　loss.　The　prediction　model　results　are　in　good　agreement　with　the　experiment　data　under　velocity　ratios　lower　than　that　of　the　design　point.　©　2016,　Science　Press.　All　right　reserved.