The design of partial multinomial regression on wear resistance of specimen was performed according to the techniques of experimental designing optimization method. Test model specimens made of 55 steel with regular concaves were processed by laser texturing technology technique. Investigations into topographical and tribological features were conducted on MG2000 wear and abrasion test machine. As one important test index, wear loss was taken to be detected by using partial polynomial regression plan, and three test factors of temperature （300 ℃, 500 ℃）, running time（180s, 300s）, rotating speed（400r/min, 600r/min）, non-smooth concave unit size（150μm, 200μm, 250μm,300μm）and unit distributing density（250μm,350μm at line and row has been fixed）were selected to investigate the effects on the surface wear loss of 55 steel samples. Multi-factor linear regression equation was derived from test data. According to their importance sorts to test results, the factors are temperature，running time, non-smooth concave unit size, rotating speed, unit distributing density respectively. Main wear-resistance mechanism of the non-smooth concave surface was analyzed at last.