Layer of Protection Analysis (LOPA) provides a simplified risk analysis technique for a select set of postulated incident scenarios, evaluating an initiating event (IE) and subsequent independent protection layers (IPLs) as a series of probabilities of failure on demand (PFDs). Human error is assessed in LOPA as either an IE or IPL. Techniques that have been recommended for human reliability assessment (HRA) by the CCPS Guidelines for IEs and IPLs in LOPA, such as THERP, are generally considered in other industries as ‘first generation’ HRA methods, meaning that they are appropriate for some instances but have their limitations. These limitations have become more apparent through the analysis of actual incidents and the understanding of the underlying human errors that have contributed to them. For example, failure to recognize the significance of high fluid returns during the negative pressure test and failure to perceive the importance of drill pipe pressure during well monitoring activities were key contributors to the Deepwater Horizon incident. These diagnosis and decision-making errors are not well addressed by first generation HRA methods and require particular timing, error mode and performance shaping factor considerations. Several HRA quantification methods are available to assess the human error probability associated with these so-called “cognitive” aspects of a task, as well as the equipment manipulation.
For HRA applications other than nuclear power, such as chemical weapons destruction plants and nuclear fuel transport and storage, other quantification methods have been successfully applied and their results accepted during peer and regulatory agency reviews. These include the Cognitive Reliability Error Analysis Method (CREAM), and the Human Error Assessment and Reduction Technique (HEART). As the name implies, CREAM focuses on the decision-making errors by providing Human Error Probabilities (HEPs) for a standard set of task types, but also includes adjustment factors to evaluate facility, organizational and scenario-specific influences. HEART contains a list of generic task types and their HEP estimates that are modified using factors associated with ‘error producing conditions’.
In each instance where HRA is applied to a new industry, an evaluation must be made regarding the degree of applicability of the standard error types, Performance Shaping Factors (PSFs) and quantification methods. Since the error types and PSFs have an underpinning in human factors, psychology and cognitive science, they generally apply regardless of the industry. For instance, a recent paper by the departments of marine/maritime engineering of two Turkish universities presented in the Journal of Loss Prevention in the Process Industries discussed the application of the CREAM cognitive error evaluation technique to the cargo loading process of LPG tankers. However, it is important to evaluate the applicability of HRA tools and techniques to ensure that industry-specific issues are being appropriately addressed.
This paper will discuss the HRA methods available for evaluating cognitive errors and their applicability to the chemical process industry.