In this paper, an improvement in the hybrid stochastic/deterministic pincus-Nelder-mead optimization algorithm (P-NMA) which enables to solve the target optimization problem of vibration-based damage detection is prop...
详细信息
In this paper, an improvement in the hybrid stochastic/deterministic pincus-Nelder-mead optimization algorithm (P-NMA) which enables to solve the target optimization problem of vibration-based damage detection is proposed. The proposed modification consists in reducing the sampling domain of the pincus formula by assigning a maximum number of damaged elements, i.e., by allowing only a few elements of the sampling vector to be different from 1. Consequently, a new parameter which determines the maximum number of damaged elements (np(max)) is introduced and must be choose by the designer. Such a modification attempts to speed up the convergence of the original version of the P-NMA and thus, reducing its computational cost. A series of numerical examples, all selected from literature, was performed. To test the accuracy and efficiency of the proposed improved optimization algorithm (IP-NMA), its results were compared to those obtained by the P-NMA and the metaheuristic harmony search algorithm (HS). A statistical analysis was also performed in order to test the robustness of the three algorithms. The proposed improved optimization algorithm showed better performance (more accurate and required lower computational cost than the original version of the P-NMA and the metaheuristic HS), emphasizing its capacity in damage diagnosis and assessment. (C) 2016 Elsevier Ltd. All rights reserved.
暂无评论