A Prevention Approach for Chemical Process Risk Management: Retrieving Knowledge from Relevant Databases through Machine Learning Techniques

Santini, Davide (2019) A Prevention Approach for Chemical Process Risk Management: Retrieving Knowledge from Relevant Databases through Machine Learning Techniques. [Laurea magistrale], Università di Bologna, Corso di Studio in Ingegneria chimica e di processo [LM-DM270], Documento full-text non disponibile
Il full-text non è disponibile per scelta dell'autore. (Contatta l'autore)


The massive amount of data produced by industrial activities is continuously increasing and represents a valuable resource. The Seveso III directive underlines the need of monitoring and analysing data to improve the safety performance of the plants. A new approach that uses data from past events that occurred in these plants, to learn from them to make predictions has been considered. In this context, machine learning techniques are suggested to retrieve knowledge from data to be able to take decisions. This work aims to suggest a new approach to manage and analyse data from different sources regarding the process industry. The knowledge retrieves, should help building and supporting prevention safety barriers, improving the overall risk management. The machine learning tools used are from the open source library TensorFlow. The three prediction models considered are a linear model, a deep model and a hybrid of the previous two. Two different levels of data source have been considered: the MHIDAS (Major Hazards Incident Data Service) database and an alarm database from a specific chemical establishment. An activity of data mining allowed preparing the databases. Fundamental differences between the developed models have been found based on their target. If it is essential to have a conservative prediction, e.g., to predict accidents events, the “recall” metric should be prioritised. On the other hand, if the goal is to improve risk awareness and promote response, the “precision” metric should be the priority. Another metric used to evaluate the performance of models is the area under the curve precision-recall. The results obtained for the two databases showed that, for the MHIDAS database, deep and hybrid models, are the most suitable to achieve good recall of accident prediction. For the alarm database, the linear model is the one that best performs for high precision to promote response and the hybrid model for a high recall of critical alarm predictions.

Tipologia del documento
Tesi di laurea (Laurea magistrale)
Autore della tesi
Santini, Davide
Relatore della tesi
Correlatore della tesi
Corso di studio
Ingegneria di processo
Ordinamento Cds
Parole chiave
machine learning,big data,industry 4.0,risk management,seveso directive,hse,tensorflow
Data di discussione della Tesi
14 Marzo 2019

Altri metadati

Gestione del documento: Visualizza il documento