The purpose of this paper is to discuss the characteristics of several stochastic simulation methods applied in computation issue of structure health monitoring (SHM).
On the basis of the previous studies, this research focusses on four promising methods: transitional Markov chain Monte Carlo (TMCMC), slice sampling, slice-Metropolis-Hasting (M-H), and TMCMC-slice algorithm. The slice-M-H is the improved slice sampling algorithm, and the TMCMC-slice is the improved TMCMC algorithm. The performances of the parameters samples generated by these four algorithms are evaluated using two examples: one is the numerical example of a cantilever plate; another is the plate experiment simulating one part of the mechanical structure.
Both the numerical example and experiment show that, identification accuracy of slice-M-H is higher than that of slice sampling; and the identification accuracy of TMCMC-slice is higher than that of TMCMC. In general, the identification accuracy of the methods based on slice (slice sampling and slice-M-H) is higher than that of the methods based on TMCMC (TMCMC and TMCMC-slice).
The stochastic simulation methods evaluated in this paper are mainly two categories of representative methods: one introduces the intermediate probability density functions, and another one is the auxiliary variable approach. This paper provides important references about the stochastic simulation methods to solve the ill-conditioned computation issue, which is commonly encountered in SHM.
The first author would like to sincerely thank Professor Jianye Ching and Dr Yi-Chu Chen in Taiwan for their selfless help of the TMCMC algorithm.
This work was supported by the National Natural Science Foundation of China (No. 50935005) and the Major State Basic Research Development Program of China (973 Program) (No. 2009CB724306). The authors express the most sincere thanks to these organizations.
Zhang, Y. and Yang, W. (2014), "A comparative study of the stochastic simulation methods applied in structural health monitoring", Engineering Computations, Vol. 31 No. 7, pp. 1484-1513. https://doi.org/10.1108/EC-07-2013-0185Download as .RIS
Emerald Group Publishing Limited
Copyright © 2014, Emerald Group Publishing Limited