NSERC’s Awards Database
Award Details

Probabilistic predictive warning system for process operations

Research Details
Application Id: RGPIN-2019-04122
Competition Year: 2019 Fiscal Year: 2019-2020
Project Lead Name: Ahmed, Salim Institution: Memorial University of Newfoundland
Department: Engineering and Applied Science, Faculty of Province: Newfoundland and Labrador
Award Amount: $28,000 Installment: 1 - 5
Program: Discovery Grants Program - Individual Selection Committee: Materials and Chemical Engineering
Research Subject: Chemical engineering Area of Application: Energy resources (including production, exploration, processing, distribution and use)
Co-Researchers: No Co-Researcher Partners: No Partners
Award Summary

This program aims to develop a probabilistic predictive approach to the design and management of early warning systems used in process industries. The special features and novelty of the proposed approach is in the use of probabilistic models for alarm annunciation and the use of risk in every stage of warning design, namely, allocation, annunciation, prioritization and management. In addition, predictions will be used for issuing warnings. ***Industries significantly benefit from using predictions for control. On the other hand, considerable resources are used for risk assessment. However, current alarm systems are not risk-based; use of predictions is also restricted. Research efforts are required to make plant alarm systems an effective tool in preventing accidents. ***We plan to develop a comprehensive early warning system based on quantified risk and prediction of events from plant data. The current plant alarm systems use a deterministic setting where abnormality is described without any reference to its likelihood. Our objective is to develop tools to describe unwanted events in terms of likelihood and perform diagnosis by considering a set of possible combinations of failures that explain the observations. Methodologies will be developed to quantify uncertainties with plant models which arise from assumptions on model structures and noises in process data. Uncertainties associated with risk models will also be developed. Correlations and dependencies among events as well as their consequences will be quantified and taken into account in the design of warning system. We also want to propose an alarm design philosophy to complement the existing standards in incorporating risk. A prototype alarm system will be developed that will be integrate-able with the existing plant alarm systems. The new system will be tested in pilot plants and industrial settings.***Process industries are in dire need of an improved alarm system. Operators are often overwhelmed by alarms and are unable to perform their main responsibilities. Risk is a measure that is well understood and highly valued. Hence, the developed methodology is expected to find widespread utility in process industries.***By developing technologies and tools that ensure smooth, accident-free operation of industrial plants, this program will contribute directly to the productivity of the industrial sector. Thus it will help to save billions of dollars in losses caused by abnormal situations in industries.***Along with the development of tools and methods, a number of highly qualified personnel will be trained in this program. The HQP will be trained on developing the tools as well as its implementation to real plants; thus the HQP will be valuable assets for the concerned industry. Finally, the developed methods will be useful for risk assessment in general with applications outside process facilities, e.g. for health monitoring, economic activities and natural disasters.***********