Problems associated with MRT LBM simulations of 3D flat Poiseuille flow

Hellow, I write a MRT-LBM code based on D3Q19 model to simulate a 3D flat Poiseuille flow, by using the Guo’s non-equilibrum extrapolation method on the pressure inlet, pressure outlet and the no-slip wall boundary. But when I finish the code to run the simulation, the velocity contour seems to be something wrong:

Have anyone done the same simulation like this? I would be very grateful if anyone could help.