Abstract
Soft sensors are mathematical models able to estimate process variables. They can work in parallel with hardware sensors, and can be implemented at a low-cost on existing hardware. They are useful for back-up of measuring devices, reduction of measuring hardware requirements, real-time estimation for monitoring and control, sensor validation, fault detection and diagnosis, what-if analysis. In industrial applications, data-driven approaches, especially based on soft-computing techniques, are very promising. In this paper we review important issues in soft sensor design and applications, especially concerning the applications in the field of nuclear fusion.

This publication has 6 references indexed in Scilit: