Open AccessMulti-objective Ensemble of Regression Chains Prediction Algorithm for Pose Correction Errors of Precise Vision-based Printing Equipment
Author(s) -
Qiuji Wu,
Zhenhui Zhan,
Shilin Yang,
Xianmin Zhang,
Lixin Yang
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1550/3/032087
Subject(s) - computer science , stability (learning theory) , algorithm , orientation (vector space) , mean absolute percentage error , position (finance) , base (topology) , set (abstract data type) , artificial intelligence , field (mathematics) , mean squared prediction error , position error , ensemble learning , machine learning , mathematics , artificial neural network , mathematical analysis , geometry , finance , pure mathematics , economics , programming language
To accurately know the working condition of the Precise Vision-based Printing Equipment (PVPE) and improve its printing accuracy and stability, this paper proposes a multi-objective ensemble of regression chains algorithm to predict the pose correction (position and orientation) errors of PVPE’s alignment platform. Since this algorithm uses XGBoost as a base learner, it’s called XGBoost Ensemble of Regressor Chain (XGB-ERC) prediction algorithm. The algorithm is verified by a test set, which is constructed based on experimental data of PVPE. The experimental result shows that the Mean Absolute Percent Error (MAPE) of this algorithm for pose correction errors x , y and θ is 6.868%, 6.495% and 5.342%, which is 25.1%, 27.6% and 16.8% higher than that of the traditional XGBoost single-objective prediction algorithm. The predicted result of the proposed algorithm can be used to compensate the correction errors of the alignment platform, which will help improve the printing accuracy and stability of PVPE, and promote the development of the field of Surface Mount Technology (SMT).