HERAX: A Microcomputer-Based Expert System Approach for Human Reliability Analysis

Abstract

The growing awareness among managers and designers that human reliability is an important determinant of risk and profitability, and the need for tools to assist in the better understanding and systematic evaluation of human reliability with the view of incorporating the results within the overall framework of the probabilistic risk assessment, have lead to the development of various human reliability analysis (HRA) techniques.

Whereas acceptability and use of the conventional HRA techniques when carrying out global probabilistic safety and reliability studies is growing, many major obstacles have restricted their widespread application, such as the high level of knowledge, judgement and experience required to select and carry out a comprehensive HRA study, and the unavailability of human experts familiar with human factors and HRA techniques.

One of the main objectives of this research project was to investigate the feasibility of using Artificial Intelligence (AI) techniques to develop an expert computer program as one possible means of overcoming some of these obstacles.

The results of this study show that the characteristics of the HRA problem domain make it ideal for the expert systems technology, and that the expert developed here has potential benefits for the analysis of human reliability, such as ease of use, and reduction of the resources and time needed for the selection and application of the appropriate HRA technique.

The expert system approach proposed here, called HERAX, is intended to assist nonhuman factors specialists, such as designers and managers, in the selection and application of the appropriate HRA technique as part of a probabilistic risk analysis. With its explanation capabilities the system could also be used as a training tool.

The actual system is based on three of the existing HRA approaches, but could be
easily extended to include more techniques. It also includes a list of the major factors that could affect human performance, as well as a data bank on human error probabilities.

Date of Award	1989
Original language	English