Expert Elicitation Methodology in the Risk Analysis of an Industrial Machine

Abstract

Calculation of the probability of occurrence of an accident involving an industrial machine such as a metal bending press requires knowledge of the failure rates. Specifically, what is needed are the failure rate of the protective device and, the failure rate associated with the human action consisting in having one’s hands between the press dies while the operator is bending a part. The first data could, in principle be obtained from the manufacturer of the device. However, in reality, this data involves knowledge of the reliability of not only the protective device but also of the associated command circuitry. In reality, such data may be difficult to obtain. Also, many important statistics relating to human performance are not collected by workplaces. So, another way to obtain the data is through expert elicitation, that is consulting people knowledgeable with the problem at hand and asking them to estimate, based on their judgement, the probabilities or failure rates that are sought. This process is often used in the literature but is seldom described in detail. In this paper, expert elicitation is used and described in order to gather relevant data for the purpose of probability estimation. Thus, eight bending press operators in a large manufacturing plant, the health and safety coordinator as well as the workers’ supervisor were solicited.

A questionnaire was handed to them consisting of a set of brief instructions followed by three questions which were provided with multiple possible qualitative probability estimates to choose from. In order to improve the quality of the probability estimates, the suggested probabilities were associated with typical accidental events which serve as a comparison basis for the participants. A general introduction was given by the author to the participants in a group meeting on the shop floor which consisted of presentation the research project, its purpose. The questions and the choice of answers were read and explained to the group. The questionnaire was then handed to them. The whole process took little time to complete. These estimates represent the experts’ estimates of the probability of occurrence of the events in question, expressed in linguistic, qualitative terms. These estimates were translated in quantitative terms through fuzzy logic technique. More specifically, a scale composed of qualitative statements and their corresponding triangular fuzzy number was established with two main simple guiding principles in mind. Firstly, the scale should reflect the probability scales found in often-used safety standards. Secondly, the fuzzy triangular numbers should not overlap so that there is no need to invert any of their components as required by the rules of fuzzy number arithmetic.