Output details
11 - Computer Science and Informatics
Lancaster University
Adaptive inferential sensors based on evolving fuzzy models
<12> This paper, published in well-respected IEEE Transactions (IF=3.6), proposes self-calibrating “evolving sensors” (called eSensors) with a focus on reducing maintenance costs in online data classification systems, while maintaining high precision and interpretability/ transparency. It introduces automatic methods for online selection of the most relevant input variables as well as for detection of shift and drift in the data stream using ageing. Results demonstrate that eSensors can be automatically derived from data streams in real time. The approach has since been applied in industry - e.g. by Dow Chemicals and in a CEPSA oil refinery in Spain.