Health Monitoring Algorithm by the Monte Carlo Filter Based on Non-Gaussian Noise

A basic study of the application of the Monte Carlo filter (MCF) to structural damage detection is reported. In that method, each probability distribution is expressed by many of its realizations, called particles or samples. The advantage of the MCF is that it deals with non-linear and non-Gaussian problems. In terms of damage detection, non-Gaussian noise may be preferable because the damage tends to be concentrated on a specific part of a structure. Two kinds of numerical examples are shown. First, the stiffness of the experimental model for the shaking table test is identified by the MCF and EKF (Extended Kalman Filter). Based on hypothetical data, numerical simulations of damage detection with non-Gaussian process noise then are performed and discussed. Because the MCF results are given by many particles (samples), the detailed probabilistic nature of the identified parameters also can be discussed.