Get access

A Novel Human Error Probability Assessment Using Fuzzy Modeling

Authors

  • Shuen-Tai Ung,

    Corresponding author
      Address correspondence to Shuen-Tai Ung, Rm. 309A, Merchant Marine Building, National Taiwan Ocean University, 2 Pei Ning Road, Keelung 20224, Taiwan; shuentai@mail.ntou.edu.tw.
    Search for more papers by this author
    • Department of Merchant Marine, College of Maritime Science and Management, National Taiwan Ocean University, Taiwan.

  • Wei-Min Shen

    Search for more papers by this author
    • Department of Merchant Marine, College of Maritime Science and Management, National Taiwan Ocean University, Taiwan.


Address correspondence to Shuen-Tai Ung, Rm. 309A, Merchant Marine Building, National Taiwan Ocean University, 2 Pei Ning Road, Keelung 20224, Taiwan; shuentai@mail.ntou.edu.tw.

Abstract

Human error is one of the significant factors contributing to accidents. Traditional human error probability (HEP) studies based on fuzzy number concepts are one of the contributions addressing such a problem. It is particularly useful under circumstances where the lack of data exists. However, the degree of the discriminability of such studies may be questioned when applied under circumstances where experts have adequate information and specific values can be determined in the abscissa of the membership function of linguistic terms, that is, the fuzzy data of each scenario considered are close to each other. In this article, a novel HEP assessment aimed at solving such a difficulty is proposed. Under the framework, the fuzzy data are equipped with linguistic terms and membership values. By establishing a rule base for data combination, followed by the defuzzification and HEP transformation processes, the HEP results can be acquired. The methodology is first examined using a test case consisting of three different scenarios of which the fuzzy data are close to each other. The results generated are compared with the outcomes produced from the traditional fuzzy HEP studies using the same test case. It is concluded that the methodology proposed in this study has a higher degree of the discriminability and is capable of providing more reasonable results. Furthermore, in situations where the lack of data exists, the proposed approach is also capable of providing the range of the HEP results based on different risk viewpoints arbitrarily established as illustrated using a real-world example.

Get access to the full text of this article

Ancillary