A novel fault prediction technique using model degradation analysis

This paper presents two practical applications where artificial neural networks have been used to solve difficult process engineering problems. Firstly, the ability of artificial neural networks to provide an accurate process model of a vitrification process is demonstrated on real process data. Vitrification is the process which encapsulates highly active liquid waste in glass to provide a safe and convenient method of storage. The second application again employs artificial neural networks, this time they are applied in a novel way in which they are used to capture non-linear system characteristics and then recalled to provide a means of detecting imminent failure of a vessel used in the vitrification process