The constraints of input and output variables commonly exist in chemical process control.The inconsistency between the system and different constraints may cause constrained model predictive controller to be infeasible，and that will bring harmful effect to practical production.Based on convex polyhedron geometry，for discrete-time state-space model of linear system，the feasibility analysis of constrained model predictive control and the reasonable constraints adjustment were discussed.An on-line algorithm about feasibility analysis and constraints adjustment was proposed that the constrained model predictive control law was solved at each step before.All of the constraints could be satisfied and control performance was perfect during the whole horizon by the algorithm.Simulation results of control problem of CSTR showed the effectiveness of the proposed algorithm.