Probabilistic Risk Assessment is being increasingly used in a number of industries such as nuclear, aerospace, chemical process, to name a few. Probabilistic Risk Assessment (PRA) characterizes risk in terms of three questions: (1) What can go wrong? (2) How likely is it? (3) What are the consequences? Probabilistic Risk Assessment studies answer these questions by systematically postulating and quantifying undesired scenarios in a highly integrated, top down fashion. The PRA process for technological systems typically includes the following steps: objective and scope definition, system familiarization, identification of initiating events, scenario modeling, quantification, uncertainty analysis, sensitivity analysis, importance ranking, and data analysis.

Fault trees and event trees are widely used tools for risk scenario analysis in PRAs of technological systems. This methodology is most suitable for systems made of hardware components. A more comprehensive treatment of risks of technical systems needs to consider the entire environment within which such systems are designed and operated. This environment includes the physical environment, the socio-economic environment, and in some cases the regulatory and oversight environment. The technical system, supported by an organization of people in charge of its operation, is at the cross-section of these environments.

In order to develop a more comprehensive risk model for these systems, an important step is to extend the modeling capabilities of the conventional Probabilistic Risk Assessment methodology to also include risks associated with human activities and organizational factors in addition to hardware and software failures and adverse conditions of the physical environment. The causal modeling should also extend to the influence of regulatory and oversight functions. This research offers such a methodology. It proposes a multi-layered modeling approach so that most the appropriate techniques are applied to different individual domains of the system. The approach is called the Hybrid Causal Logic (HCL) methodology. The main layers include: (a) A model to define safety/risk context. This is done using a technique known as event sequence diagram (ESD) method that helps define the kinds of accidents and incidents that can occur in relation to the system being considered; (b) A model that captures the behaviors of the physical system (hardware, software, and environmental factors) as possible causes or contributing factors to accidents and incidents delineated by the event sequence diagrams. This is done by common system modeling techniques such as fault tress (FT); and (c) A model to extend the causal chain of events to their potential human and organizational roots. This is done using Bayesian belief networks (BBN). Bayesian belief networks are particularly useful as they do not require complete knowledge of the relation between causes and effects. The integrated model is therefore a hybrid causal model with the corresponding sets of taxonomies and analytical and computational procedures.

In this research, a methodology to combine fault trees, event trees or event sequence diagrams, and Bayesian belief networks has been introduced. Since such hybrid models involve significant interdependencies, the nature of such dependencies are first determined to pave the way for developing proper algorithmic solutions of the logic model. Major achievements of this work are: (1) development of the Hybrid Causal Logic model concept and quantification algorithms; (2) development and testing of computer implementation of algorithms (collaborative work); (3) development and implementation of algorithms for HCL-based importance measures, an uncertainty propagation method the BBN models, and algorithms for qualitative-quantitative Bayesian belief networks; and (4) development and testing of the Integrated Risk Information System (IRIS) software based on HCL methodology.