ABSTRACT
Historically the rate by human error on a task were raised through expert opinion on task suggesting possible failure rates for their experience and that later were statistically adjusted to represent the best value. Other contextual variables (brightness, noise, time pressure, etc.) and individual (temperament, experience, character) were used to adjust the result to a specific reality. In this complex situation a specific value is not satisfactory, because there is no single mathematical solution, but several solutions of scenarios. One proposed solution, originated from the processing capacity arising with computers, was the introduction of networks of variable analysis (as baysean models, cognitive maps or multi-variable analyzes), and the use of fuzzy descriptors to represent the information coming from evaluations to review non-numerical variables. Regardless of the modeling tools used, this study proposes a new way of analyzing the Human Reliability, basing on the approach to interpret the failure mechanisms of human relations not as direct cause and effect but as an interpretation that success and human failure have the same physical and mental processes. The methodology references the classic techniques still used and accepted in risk analysis and reliability, as THERP, SLIM, HEART, MER-MOS and CREAM, starting from a task analysis, considering the contextual Factors that Influence the Human Performance and judged by experts using units of physical and mental task performance (actions). For demands where the risk propensity is desired, only a relative value is used, serving for the actions of improvements. The proposed solution allows for dynamic assessments to absorb in modeling how risk and safety are affected over time and context. This allows capturing the dynamic nature of the decisions that can dramatically change the course of an event. The developed method is applied to tasks of operators Plants Generation of hydroelectric power, and the context of several plants of the same company in Brazil. This application is part of a research project funded by ANEEL—Brazilian Electricity Regulatory Agency.
