Localization Method of Loose Particles Based on Chaos Theory and
Particle Swarm Optimization-Back-Propagation Neural Network 01-15-02-0012
This also appears in
SAE International Journal of Aerospace-V131-1EJ
Loose particles inside the additional pipe of a rocket engine are an important
factor that causes propulsion system failure. For loose particles inside the
additional pipe, it is necessary not only to determine whether they exist or
not, but also to locate them for subsequent processing. Due to the complex
structure of the additional pipe, the uneven medium used for sound wave
transmission, and the anisotropic speed of the sound. Thus, it is difficult to
determine the locations of loose particles by using the traditional time
difference localization method. Aiming at this problem, this article proposed a
localization method of loose particles based on Chaos Theory and Particle Swarm
Optimization-Back-Propagation Neural Network (PSO BP Neural Network). First,
chaotic characteristics of collision signals generated by loose particles are
studied. On this basis, the localization method of loose particles based on PSO
BP Neural Network is proposed, which uses the correlation dimension, Lyapunov
exponent, and the Kolmogorov entropy (K entropy) as localization features. The
test results show that the proposed loose particle localization method can
effectively locate loose particles inside a section of broken line pipe, which
is composed of composite materials and have a certain internal structure. The
method can theoretically be applied to the localization of collision signals
with similar generation mechanism.
Citation: Sun, Z., Wang, G., Gao, M., Gao, Y. et al., "Localization Method of Loose Particles Based on Chaos Theory and Particle Swarm Optimization-Back-Propagation Neural Network," SAE Int. J. Aerosp. 15(2):185-196, 2022, https://doi.org/10.4271/01-15-02-0012. Download Citation
Affiliated:
Heilongjiang University, Electronic Engineering College, China Harbin Institute of Technology, Reliability Institute for Electric
Apparatus and Electronics, China
Pages: 12
ISSN:
1946-3855
e-ISSN:
1946-3901
Related Topics:
Neural networks
Rocket engines
Nanomaterials
Composite materials
Failure analysis
SAE MOBILUS
Subscribers can view annotate, and download all of SAE's content.
Learn More »