Risk-time Risk Tracker through mapping HAZOP into Bayesian network using machine learning

The first step in analyzing potential risks in a process system is called hazard identification. As the name states, it identifies the potential danger in relation to system operations. In other words, this step answers what can go wrong in a system. This results in a listing of what can go wrong, its causes and consequences. There are established techniques that were developed in the past for this purpose. The industry has been using them for over half of a century. Since then, the complexity of the systems has significantly increased. This project aims to upgrade one of the most widely used techniques in hazard identification, called HAZOP. HAZOP is a simple table with text listing all the potential hazards along their causes, consequence, existing safeguards and recommendations. The text is static in nature and usually revalidated every five years. In between, there are a few changes that were not considered, which makes the existing HAZOP outdated and may cause inaccurate or misleading results. This project aims to transform HAZOP from a simple text into a machine learning (ML) algorithm continuously fed by real-time parameters to provide a real-time tracking of known and unknown risk in relation to system operation.

Intern: 
Mohammed Taleb Berrouane
Faculty Supervisor: 
Carlos Bazan
Province: 
NL
Partner University: 
Discipline: