Uncertainty Characterization of Engineering Failure Data

Abstract

How to quantitatively measure the uncertainty of engineering failure data is an important but still unsolved task in probabilistic risk analysis. This paper aims to fill the gap first by specifying the requirements for a robust uncertainty measure to meet the criteria. Complexity and uncertainty measurements in computational complexity, classical statistical mechanics and information theory are also reviewed for possible inspiration. In this paper, a new groundbreaking parameter, which is related to reliability or survival function, is selected to characterize the uncertainty of engineering failure data with given probabilistic distributions. The uncertainty formulae based on the Shannon entropy and the new uncertainty parameter for various distribution functions are also provided. Finally, several examples are given to demonstrate the applicability of the new uncertainty measure in durability and reliability analyses.