Solving inverse problems using Particle Swarm Optimization: An application to aircraft fuel measurement considering sensor failure

Abstract

This paper describes a robust modeling method to handle inverse problems with missing data. The modeling method is applied to aircraft fuel measurement considering sensor failure. Neural Networks that are tolerant to noisy data are adapted to approximate the nonlinear physical process. Unlike previous algorithms that use gradient information to search input space in inverse problems, the proposed method thoroughly explores the input space using particle swarm optimization. The comparison results show the effectiveness of our method in dealing with missing data.