|
PRIOR KNOWLEDGE INTEGRATION IN SELFORGANIZING MAPS FOR COMPLEX PROCESS SUPERVISION
Ignacio Díaz Blanco* Alberto Benjamín Diez González*
Abel Alberto Cuadrado Vega* Manuel Domínguez González**
* Universidad de Oviedo.
Departamento de Ingeniería Eléctrica, Electrónica, de Computadores y
Sistemas. Área de Ingeniería de Sistemas y Automática. Campus de
Viesques s/n, 33204, Gijón, Asturias, Spain
** Universidad de León.
Departamento de Ingeniería Eléctrica y Electrónica. Ingeniería de
Sistemas y Automática. Escuela de Ingenierías Industrial e Informática
Edificio Tecnológico Campus de Vegazana, 24071, León, Spain
The SOM allows to project high dimensional feature vectors of a process on a low dimensional (2D) space, where the process state trajectory can be displayed for supervision and visualization. Regions in this space describe different process conditions. This paper suggests three novel methods to define regions in the visualization space associated to different process conditions using prior knowledge of the process given as: a) Labeled data sets, b) Process model, and c) Fuzzy rules
Keywords: State Monitoring, Fault Identification, Neural Networks, Fuzzy Inference, Supervision
Session slot T-Mo-A08: Decision Support & Process Supervision/Area code 7e : Fault Detection, Supervision and Safety of Technical Processes
 | 