Soft sensor design with small datasets using a difference-based neural network

doi:10.1016/j.engappai.2026.115734

Soft sensor design with small datasets using a difference-based neural network

2026-07-23

0

OA

AI

Maria Gabriella Xibilia

DOI：10.1016/j.engappai.2026.115734

Original

Original request for help

Save

Abstract

En 中文

• The Δ -Net is proposed to address dynamic system identification from small datasets. • The Δ -Net uses an augmented dataset obtained by a combinatorial approach. • Real industrial benchmarks are used to validate the proposed method. • Comparisons with different baseline methods show the superior performance of Δ -Net. • Statistical analyses for different dataset dimensions are reported.

Keywords:

Dynamical system identification

Soft sensors

Small datasets

Data augmentation

Neural networks

Difference-based learning

Soft sensor design with small datasets using a difference-based neural network

Abstract

Journal

Organization