Data driven methods for decision dupport and monitoring

Description of impact

Description: Efficient exploitation of historical data contained in data bases following data mining and knowledge discovery methodology to build decision support systems (DSS) and to extend monitoring system with fault detection and diagnosis capabilities. Main techniques addressed in this research line include: - Multivariate Statistical Process Control methods, as PCA and PLS and its application to process understanding and fault detection (control charts) and fault diagnosis (contributions analysis) of both continuous and batch processes. - Sensors Reconstruction and Soft sensors: machine learning and statistical methods to exploit redundancy from multiple sources and to predict unknown variables. - Qualitative representation of signals and trends. - Structured (sequences and graphs) data mining: Sequence pattern discovery algorithms and their application to fault forecasting, similarity analysis among graph based entities - Case based reasoning (CBR) for supporting decisions according to previous experiences. - Complex Event Processing (CEP). Application domains: - Power Systems: power quality monitoring, energy efficiency monitoring, fault location in distribution systems. - Medical and healthcare systems: decision support to intervention, clinical and familiar data modelling, support to diagnosis, population models. - Industry: chemical process indu plastic injection, aerospace engines. - Environment: waste water treatment plants.