Reliability-Based Design of Human Performance Conditions Using Fuzzy Perfection
- PDF / 173,192 Bytes
- 13 Pages / 594 x 792 pts Page_size
- 53 Downloads / 174 Views
RELIABILITY-BASED DESIGN OF HUMAN PERFORMANCE CONDITIONS USING FUZZY PERFECTION
UDC 681.5.015:007
A. Rotshtein
Abstract. A method is proposed for selection of performance conditions that affect human reliability without time-consuming calculation of the probability of human error. This method is based on the specially introduced concept of fuzzy perfection and theory of decision-making under fuzziness. It is shown that the proposed method can be used both independently and together with the well-known CREAM method of determining the reliability class based on cognitive assessments of human performance conditions. It is demonstrated that the results obtained by the proposed fuzzy perfection method coincide with the ones based on probabilities of erroneous actions. Keywords: performance conditions, human error probability, fuzzy logic inference, fuzzy perfection, intersection of fuzzy criteria. INTRODUCTION Human errors are reasons of many accidents in transportation, industrial, and other human-machine systems. Their probability depends on factors that determine human performance conditions: workplace organization, professionalism, work intensity, exhaustion, etc. Optimal choice of performance conditions that minimize human error probability is the most important part of human-machine system design process. To formulate this problem in terms of classical mathematical programming, it is necessary to define the following: — controlled variables, which are human performance conditions; — objective function, which relates human error probability with assessments of human performance conditions; — constraints on admissible values of performance conditions corresponding to the available resources. The main difficulties in implementing such approach are due to the following: (i) experts assess many human performance conditions from the point of view of quality and (ii) the domain of admissible values of these estimates may not have an exact boundary. Because of these difficulties, the trial and error method accepted in systems analysis [1] is widely applied in design practice: a proposed variant is assessed, then improvements are introduced, and it is assessed once again, etc. Nonlinear dependence of error probability on contributing factors and necessity of handling expert information stipulate the use of the apparatus of fuzzy rules in modeling [2], which are a universal approximator of nonlinear functions. The fuzzy CREAM method [3, 4], which is used for reliability analysis of human performance in increased-risk systems (transportation, nuclear power engineering, etc.) implement such approach to the fullest extent. The fuzzy CREAM method is based on the basic CREAM method [5], which determines error probability class from the number of performance conditions influencing increase and decrease in the reliability of human performance. Jerusalem College of Technology, Machon Lev, Israel, [email protected]. Translated from Kibernetika i Sistemnyi Analiz, No. 2, March–April, 2019, pp. 82–95. Original article submitted
Data Loading...
