Making　quality　control　in　the　design　stage　can　greatly　improve　product＇s　robustness　and　save　costs.　The　whole　system　function　variance　model　and　the　sensitivity　index　model　are　established　combined　with　axiomatic　design　and　then　making　comparative　evaluation　to　them　combined　with　information　axiom.　A　fundamental　model　of　describing　relationships　among　function　and　structural　characteristics,　design　parameters　and　uncontrollable　factors　in　non-linear　systems　which　based　on　Taylor　expansion　is　proposed.　And　then　the　relation　between　robust　design,　information　axiom　and　the　concept　of　information　is　expounded.　The　solution　to　information　amounts　can　be　used　as　the　evaluation　index　to　the　result　of　robust　design.　Finally,　combined　with　coating　head　example　show　that　the　method　can　optimize　design　parameters　reasonably,　reduce　design　time　and　cost　effectively　under　the　premise　of　meeting　product　performance.