Based on a two-dimensional Poiseuille and Wormersley flow, accuracy and grid convergence of velocity, shear stress and wall shear stress (WSS) measurements were investigated using the single-relaxation-time (SRT) and multiple-relaxation-time (MRT) lattice Boltzmann models under various open and wall boundary conditions. The results showed that grid convergence of shear stress and WSS are not consistent with that of velocity when flow channels are not aligned to the grids, and strongly rely on the used wall boundary conditions. And the MRT model is slightly superior to the SRT when simulating the complicated border flow. Moreover the WSS should be approximately calculated on the fluid nodes closest to walls under the curved boundary (CB) condition but not for the bounce-back (BB) boundary scheme. As applications, distributions of WSS in a wavy-walled channel and distensible carotid artery were simulated which would much more depend on local roughness of vessel intima than channel diameters.