Summary:
UCLA researchers in the Garrick Institute for the Risk Sciences have developed a software for performing human reliability analyses using a model-based method, which provides more consistent, traceable, and reproducible results.
Background:
Statistical data regarding accidents in complex person operated technical systems indicate a significant human contribution to these incidents. A substantial proportion—between 60% to 90%—of system failures can be attributed to human error. Human Reliability Analysis (HRA) has progressed to better account for factors influencing unsafe human actions and thus offer improved estimates for the likelihood of human error in probabilistic risk assessments (PRAs). First-generation HRA methods were introduced to enhance PRAs but suffered from limitations such as a lack of causal understanding of operator errors and inconsistency in results due to insufficiently structured methods. The development of second-generation HRA methods places greater emphasis on operator awareness and context, yet they still rely on subjective methodologies and expert-derived probabilities. Collectively, these issues found in both first- and second-generation HRA methods have led to inconsistencies, insufficient traceability, and a lack of reproducibility in both qualitative and quantitative phases. These challenges have led to significant variability in outcomes across diverse applications of HRA methods, including cases where different analysts use the same method. As a result, there is a pressing need to invent a tool that can facilitate the attainment of dependable and uniform outcomes in HRAs.
Innovation:
Professor Ali Mosleh and his colleagues have developed a novel software that employs a model-based approach for conducting HRAs, ensuring greater consistency, traceability, and reproducibility in the analysis process. It employs an innovative strategy known as the Phoenix HRA methodology, specifically crafted to address challenges encountered within the realm of HRA. Built upon a cognitive model of human response, Phoenix merges robust components of established HRA best practices, incorporates insights from empirical studies, and draws from the strengths of both established and emerging HRA methodologies. This approach also offers a framework for encompassing the contextual factors linked to Human Failure Events (HFEs), encompassing errors arising from both omission and commission. It incorporates relevant insights from cognitive psychology literature and real-world operational experiences to identify potential sources of failure and the influences impacting crew-plant interactions under procedure-driven and knowledge-supported conditions. In comparison to existing software solutions for HRA, this tool emerges as a preferred alternative, providing a more comprehensive and effective tool for performing HRA. Moreover, this versatile software allows for seamless expansion into diverse PRA domains based on its foundational principles, such as nuclear, oil and gas, aerospace, aviation, and healthcare environments.
Potential Applications:
- Accident sequence precursor analysis for complex systems, such as:
- Nuclear power plants (NPPs)
- Space missions
- Transportation networks, etc.
- Event assessment for regulatory compliance in healthcare and pharmaceutical industries
- Power operations and shut down operations for NPPs and manufacturing plants
Advantages:
• Traceability
• Reproducibility
• Consistency
• Realistic context depiction of human error
• Enhanced interdependency treatment
Development to Date:
Approximate Year(s) of Public Disclosures of Software: 2020.
Reference:
UCLA Case No. 2023-088
Lead Inventor:
Prof. Ali Mosleh