Journal article
Efficient Filter Generation Based on Particle Swarm Optimization Algorithm
IEEE access, Vol.9, pp.22816-22823
2021
DOI: 10.1109/ACCESS.2021.3056464
Abstract
The evaluation of Hankel integration is an important part in the interpretation of electromagnetic (EM) data, especially in physical and geophysical applications. The digital linear filter (DLF) method is commonly applied. For an optimal digital filter design based on matrix inversion, it requires optimization over the model space of the spacing and shift. This is typically obtained by a grid search algorithm on gradually refined grids. In this paper, we apply a particle swarm algorithm to optimize the spacing and shift in the model space. In this algorithm, we use a group of particles to search the model space and do not have to grid the model space sequentially to finer meshes. It has been applied to search the optimal spacing and shift for analytical function pairs, indicating a fast convergence. The performance of the obtained 201-point filter is examined and compared with other published filters based on analytic function pairs and controlled source electromagnetic (CSEM) applications. The results indicate that the proposed 201-point filters have a good numerical performance in terms of accuracy over a large offset range.
Details
- Title: Subtitle
- Efficient Filter Generation Based on Particle Swarm Optimization Algorithm
- Creators
- Liang Zeng - Xiamen UniversityJintai Li - Central South UniversityJianxin Liu - Central South UniversityRongwen Guo - Central South UniversityHang Chen - Central South UniversityRong Liu - Central South University
- Resource Type
- Journal article
- Publication Details
- IEEE access, Vol.9, pp.22816-22823
- DOI
- 10.1109/ACCESS.2021.3056464
- ISSN
- 2169-3536
- eISSN
- 2169-3536
- Publisher
- IEEE
- Number of pages
- 8
- Grant note
- 41904158; 42004065; 42074165 / National Natural Science Foundation of China (10.13039/501100001809)
- Language
- English
- Date published
- 2021
- Academic Unit
- Earth and Environmental Sciences
- Record Identifier
- 9984962626402771
Metrics
4 Record Views