Integrating transfer learning within data-driven soft sensor design to accelerate product quality control

• DOI: 10.1016/j.dche.2024.100142
• Publication Date: 2024-01-26T12:22:06Z
• Authors: Sam Kay; Harry Kay; Max Mowbray; Amanda Lane; Cesar Mendoza; Philip Martin; Dongda Zhang
• Keywords: Soft sensor; Benchmark (surveying); Transfer of learning

The measurement of batch quality indicators in real time operation is plagued with many challenges, hence soft sensing has become a promising solution within industrial research. However, small data traditionally been severe problem, hindering the ability to create accurate, reliable sensors, especially research and development for new product formulations. Nevertheless, it often case that modelling knowledge available related system. In order exploit this, we have developed generalisable transfer learning methodology which takes advantage previous efforts accelerate improve construction models systems. Specifically, adapted recently advanced data-driven made an existing process formulation integrated feature-based approach facilitate two systems, each containing notable differences original. performance sensors was tested rigorously compared benchmark under different availability conditions. It shown that, proposed mechanism yielded high accuracy, robust scenarios, indicating its potential use novel