This research activity aims to use the data made available by the process monitoring system for its safety. On the one hand, the possibility of developing soft-sensors, based on neural networks, will be exploited to monitor variables that are not easily accessible via hardware sensors, but are of interest for running the process under safe conditions. Besides, the availability of a model based on neural networks as a surrogate for a first principles model allows the use of advanced centralized control methodologies, for example of the Model Predictive Control system. Furthermore, always based on the data obtained from process monitoring, algorithms are developed to identify process anomalies deriving from internal and external process disturbances, as cyberattacks, thus allowing for a more rapid risk management and avoiding critical events onset.
