A 3-DOF bionic elbow joint actuated by Pneumatic artificial muscle (PAM) was designed in this paper, and its inverse kinematics model was also established. Then, based on the Model-free adaptive control (MFAC) theory and the effects of control parameters to the control system, a Parameter self-adjust Model-free adaptive control (PSA-MFAC) strategy was proposed, and its adaptability for different control objects was also tested in simulation environment. Combined with the inverse kinematics model, motion control experiments of the bionic elbow joint were conducted in semi-physical platform. Compared with conventional MFAC and PID control algorithm, the experiment results strongly verified the improvement of PSA-MFAC control accuracy. The tracking accuracy of conventional MFAC and PID controller were 9.5 % and 15 %, respectively, in contrast, the PSA-MFAC controller was only 3.8 %. Moreover, complex dynamics modelling of the elbow joint and adjusting process of control parameters were neglected in PSA-MFAC control system.