Intelligent prediction and decision making in safety operations and process engineering

  • Favour Chidinma Ikwan

Student thesis: Doctoral Thesis

Abstract

This research created a new predictive decision making system and investigated ways to alleviate risks in process engineering. The system was based on trend analysis of historical data. It predicted the probable future condition of trait ‘states’ and potential situations that may lead to catastrophe.
Initially, the work investigated the state of the art and literature, and gaps in the research were identified. A gap in the research was for a prediction and decision making system to monitor data, and predict potential situations that may lead to catastrophic failures. The work moved to concentrate on decision-making and ways to alleviate risks using a more systematic approach. That idea was presented to the Petroleum Institute, and they approved of the direction of the work. Initially, a fault tree analysis was conducted, and a new fault tree model of a storage tank was created based on a model created by Painting. The new model was tested with pseudo-random data that covered all reasonable eventualities. A Fault tree model for a storage tank was created and coded within Excel. Information on current ‘state’ was shown on a coloured 2-D fault tree model in Excel, showing a traffic light system of green, yellow, amber, and red to visualise information.
A prototype intelligent decision making system was created using a dynamic risk model that combined real-time data and historical data to show real-time risk likelihood. The dynamic model could aid a decision maker to quantify, aggregate, and understand the current risk when making decisions. The prototype decision making system could only show the current state. It could not help a decision maker predict risk and did not provide prioritized and preventive measures.
The work moved to investigate other decision making methods including Best Worst Method, PROMITHEE I & II, Analytical Hierarchy Process (AHP) and artificial intelligent methods. These were compared to select a suitable method to evaluate risks in process engineering. A new decision making method that combined valuable aspects of AHP and PROMITHEE and artificial intelligence (AI) was created that provided prioritized and preventive measures. The AI included the use of time series to forecast data and rule-based methods to prioritise outputs.
The parameters and variables for the system were established. Then the algorithms were produced, and the system tested. The system was based upon lessons learnt from historic catastrophic incidents. The new system considered a petroleum storage tank using a combination of data called traits. It predicted the traits likely to cause a catastrophic failure and could improve safety and reduce down time and might be used to prioritise funding.
Date of Award2022
Original languageEnglish
SupervisorDavid Sanders (Supervisor), Giles Tewkesbury (Supervisor) & Mohamed Galal Hassan Sayed (Supervisor)

Cite this

'