Neural networks with upper and lower bound constraints and its application on industrial soft sensing modeling with missing values

作者：

Highlights：

• The overestimation of the prediction error caused by filling methods is analyzed.

• The neural network framework with upper and lower constraints is carried out.

• Mask-specific local models are use to improve the accuracy of the bounds.

• The objective goal maximizes the posterior probability under bound constraints.

摘要

•The overestimation of the prediction error caused by filling methods is analyzed.•The neural network framework with upper and lower constraints is carried out.•Mask-specific local models are use to improve the accuracy of the bounds.•The objective goal maximizes the posterior probability under bound constraints.

论文关键词：Neural network,Upper and lower bound,Missing value,Soft sensing

论文评审过程：Received 18 May 2021, Revised 18 November 2021, Accepted 26 February 2022, Available online 7 March 2022, Version of Record 17 March 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.108510