Transfer learning-based thermal error prediction and control with deep residual LSTM network

作者：

Highlights：

• Mechanism- and data-models are combined to improve accuracy.

• New variable step is designed to propose ILMS.

• DSHOA is proposed to optimize hyper-parameters of DRLSTMN network.

• ILMS-DSHOA-DRLSTMN is proposed to improve predictive accuracy.

• Transfer learning model is trained to improve robustness.

摘要

•Mechanism- and data-models are combined to improve accuracy.•New variable step is designed to propose ILMS.•DSHOA is proposed to optimize hyper-parameters of DRLSTMN network.•ILMS-DSHOA-DRLSTMN is proposed to improve predictive accuracy.•Transfer learning model is trained to improve robustness.

论文关键词：Machine tool,Thermal error,Error compensation,Temperature rise,Spindle system

论文评审过程：Received 18 April 2021, Revised 1 November 2021, Accepted 5 November 2021, Available online 20 November 2021, Version of Record 10 January 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107704