.png){kind=icon}
.png){kind=icon}
Return
A compensation approach for magnetic encoder error based on improved deep belief network algorithm
.png){kind=icon}
PRE
AI
DOI：10.1016/j.sna.2023.115003.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Magnetic encoders are an important part of industrial automation control systems and are widely used in industrial production. In this paper, a magnetic encoder is designed for angle detection during robot arm motion, but the accuracy is not high in practical applications. Through experiments, it is found that the main error compensation methods at present cannot effectively improve the accuracy of this encoder. Therefore, a hybrid prediction model based on decomposition strategy and deep learning prediction method is proposed in this paper. The hybrid prediction model structure is divided into two main parts: feature engineering based on the variational modal decomposition (VMD) method and a deep belief network prediction model based on particle swarm optimization. Through theoretical analysis, this paper introduces temperature into the prediction feature sequence, and effectively reduces the interference of errors on the prediction results through feature engineering. Experiments prove that the proposed model has excellent compensation effect, improving the accuracy from 0.22 degrees to 0.0025 degrees. The RMSE, Max_error and accuracy (sigma) of the proposed model are optimal when compared with the mainstream error compensation methods, such as long short-term memory networks (LSTM), support vector machines (SVM) and deep belief networks (DBN). This proves that the proposed hybrid prediction model has a great improvement in the compensation effect of this magnetic encoder. Our work provides a promising approach that can provide unparalleled value in improving the accuracy of magnetic encoders.
Keywords:
Magnetic encoder
Error compensation
Deep belief network
Variational mode decomposition
Particle swarm optimization
Journal
IF:
4.9
Papers:
1.5W
Citations:
3.3W
View more organizations