A new evidence-theory-based method for response analysis of acoustic system with epistemic uncertainty by using Jacobi expansion

Abstract

Evidence theory has strong ability to handle epistemic uncertainties whose precise probability distributions cannot be obtained due to limited information. However, the excessive computational cost produced by repetitively extreme value analysis severely influences the practical application of evidence theory. This paper aims to develop an efficient algorithm for epistemic uncertainty analysis of acoustic problem under evidence theory. Based on the orthogonal polynomial approximation theory, a numerical approach named as the evidence-theory-based Jacobi expansion method (ETJEM) is proposed. In ETJEM, the response of acoustic system with evidence variables is approximated by Jacobi expansion, through which the repetitively extreme value analysis needed in evidence theory can be efficiently performed. The parametric Jacobi polynomial of Jacobi expansion holds a large number of polynomials as special cases, such as the Legendre polynomial and Chebyshev polynomial. Thus, the ETJEM permits a much wider choice of polynomial bases to control the error of approximation than the traditional evidence-theory-based orthogonal polynomial approximation method, in which only the Legendre polynomial is used for approximation. Three numerical examples are employed to demonstrate the effectiveness of the proposed methodology, including a mathematic problem with explicit expression and two engineering applications in acoustic field. In these three numerical examples, efficiency and accuracy are fully studied by comparing with Legendre expansion method as well as Monte Carlo simulations.