Enhancing Spindle Precision: Thermal Error Modeling with Multi-parameter Optimization and Energy Consumption Data

Thermal error prediction models for key components in precision computer numerical control machine tools often fail to maintain high accuracy under varying working conditions, significantly impairing the effectiveness of error compensation. Furthermore, existing models frequently disregard correlati...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:International journal of precision engineering and manufacturing Ročník 26; číslo 8; s. 1837 - 1853
Hlavní autoři: Zhou, Bo, Chen, Guo-hua, Mao, Jie, Li, Bo, Fu, Zhen-xin, Li, Tao
Médium: Journal Article
Jazyk:angličtina
Vydáno: Seoul Korean Society for Precision Engineering 01.08.2025
Springer Nature B.V
Témata:
Accuracy
Adaptive algorithms
Algorithms
Boundary conditions
Confidence intervals
Density
Design
Energy consumption
Engineering
Error compensation
Experiments
Heat
Industrial and Production Engineering
Machine tools
Materials Science
Neural networks
Numerical controls
Optimization
Parameter modification
Prediction models
Real time
Regular Paper
Sensors
Software
Spindles
Working conditions
MBSCPO
Energy consumption data
Electric spindle
Thermal error modeling method
Prediction interval
ISSN:2234-7593, 2005-4602
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Thermal error prediction models for key components in precision computer numerical control machine tools often fail to maintain high accuracy under varying working conditions, significantly impairing the effectiveness of error compensation. Furthermore, existing models frequently disregard correlations between thermal errors and non-temperature-related variables. Therefore, an electric spindle thermal error modeling method based on energy consumption data and a multi-parameter optimization algorithm is proposed in this study. Based on energy consumption data(current, power), monitoring of eight key temperature points, and real-time thermal error data, a network model integrating the Bidirectional Inception-based Temporal Convolutional Network, the Bidirectional Gated Recurrent Unit (BIGRU), and the multi-head attention mechanism is established. The model's hyper-parameters and kernel density are dynamically optimized through the Modified Crested Porcupine Optimization algorithm and the Adaptive Kernel Density Estimation method. Through practical verification under variable working conditions (covering different rotational speeds), compared with the BIGRU model, the coefficient of determination (R 2 ) of the proposed model is increased by 25%. Meanwhile, the Prediction Interval at 95% confidence level with Asymmetric Width of the proposed model is 23%, highlighting the superiority of this modeling method in terms of robustness and generalization ability.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2234-7593
2005-4602
DOI:10.1007/s12541-025-01230-9