In　this　paper,　experimental　factors　are　decided　which　influence　microirregularities　Ry,　Sm　on　ground　bearing　race　surface.　The　multiple-factor　experiments　are　done,　based　on　the　uniformity　experiment　design　table,　in　order　to　make　the　experiments　overall　reflect　the　effects　of　grinding　parameters　on　microirregularity　characteristic　parameters　R　y,　Sm　of　ground　surface　with　as　few　experiments　as　possible.　A　neural　network　model　is　established　with　the　grinding　experimental　data,　which　relates　characteristic　parameters　Ry,　Sm　of　ground　surface　microirregularities　to　the　grinding　parameters,　by　means　of　the　terminal　attractor　based　BP　learning　algorithm.　Based　on　the　neural　network　model　for　the　characteristic　parameters　Ry,　Sm　of　ground　surface,　some　valuable　conclusions　have　been　drawn　by　further　multidimensional　analysis.