Kinematic error model of a 3-PUU parallel mechanism for translational motion

[Display omitted] •A revised kinematic error model, link-orientation-based joint error model (LOBJEM), is proposed.•The sensitivity of each error term in the working space is analyzed.•The compensation result of LOBJEM is better than the original error model.•A backpropagation neural network is cons...

Full description

Saved in:
Bibliographic Details
Published in:Measurement : journal of the International Measurement Confederation Vol. 202; p. 111853
Main Authors: Ke, Jian-Dong, Wang, Yu-Jen, Tsai, Jhy-Cherng
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.10.2022
Subjects:
Backpropagation neural network
Error estimation
Kinematic error model
Parallel kinematic machine
Vertically distributed 3-PUU PKM
Parallel kinematic machine
Error estimation
Backpropagation neural network
Kinematic error model
Vertically distributed 3-PUU PKM
ISSN:0263-2241, 1873-412X
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:[Display omitted] •A revised kinematic error model, link-orientation-based joint error model (LOBJEM), is proposed.•The sensitivity of each error term in the working space is analyzed.•The compensation result of LOBJEM is better than the original error model.•A backpropagation neural network is constructed to estimate the residual error caused by NGPs.•The final remained error is reduced to 0.03 mm. Accuracy of parallel kinematic machine (PKM) is affected by geometric parameters (GPs) and non-geometric parameters (NGPs). For GPs, the use kinematic error model (KEM) to calibrate these errors has been studied, but few of them have considered the influence of the linkage posture. In this paper, link-orientation-based joint error model (LOBJEM) and sensitivity analysis are presented. In LOBJEM, the original KEM, which is constant joint error model (CJEM), is derived first. The angle between each link and the Z axis is considered as the parameter to adjust the KEM to derive LOBJEM and it can cause the value of joint error to vary with the position of end effector. The calibration process includes measuring the positioning error by a laser interferometer, identifying the error terms using least square method and compensating by the numerical control code (NC-code). The error model and calibration process have been realized on the delta robot 3D printer. Compared with the CJEM, the residual error of LOBJEM was reduced from 75% to 60% of the original error. To estimate the positioning errors caused by NGPs, we constructed a backpropagation neural network model, which can reduce the final residual error to 25% of the original error.
ISSN:0263-2241
1873-412X
DOI:10.1016/j.measurement.2022.111853