Positioning error estimation models for horizontal-distributed PUU parallel mechanism

Abstract

In this paper, two models were proposed to estimate the positioning error of a 3-translational prismatic-universal-universal parallel kinematic mechanism. The two models were a kinematic error model (KEM) and a backpropagation neural network (BPNN) model, respectively. The KEM was constructed by incorporating three translational joint errors into the ideal kinematic model to describe the errors that occur during machining or assembly. Additionally, a sensitivity analysis was presented for each error parameter. The BPNN model was constructed to establish the relationship between the position of the end effector, the posture of each link, and the positioning error of the end effector using a neural network approach. Moreover, a hybrid method was proposed to decrease the final estimated residual error. The average errors of the KEM and BPNN model were 35% and 15% of the original error, respectively. The hybrid model reduced the final average error to less than 10% of its original value.