Toggle Main Menu Toggle Search

Open Access padlockePrints

Defect Profile Estimation from Magnetic Flux Leakage Signal via Efficient Managing Particle Swarm Optimization

Lookup NU author(s): Ping Wang, Professor Gui Yun TianORCiD



This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).


In this paper, efficient managing particle swarm optimization (EMPSO) for high dimension problem is proposed to estimate defect profile from magnetic flux leakage (MFL) signal. In the proposed EMPSO, in order to strengthen exchange of information among particles, particle pair model was built. For more efficient searching when facing different landscapes of problems, velocity updating scheme including three velocity updating models was also proposed. In addition, for more chances to search optimum solution out, automatic particle selection for re-initialization was implemented. The optimization results of six benchmark functions show EMPSO performs well when optimizing 100-D problems. The defect simulation results demonstrate that the inversing technique based on EMPSO outperforms the one based on self-learning particle swarm optimizer (SLPSO), and the estimated profiles are still close to the desired profiles with the presence of low noise in MFL signal. The results estimated from real MFL signal by EMPSO-based inversing technique also indicate that the algorithm is capable of providing an accurate solution of the defect profile with real signal. Both the simulation results and experiment results show the computing time of the EMPSO-based inversing technique is reduced by 20%-30% than that of the SLPSO-based inversing technique.

Publication metadata

Author(s): Han WH, Xu J, Wang P, Tian GY

Publication type: Article

Publication status: Published

Journal: Sensors

Year: 2014

Volume: 14

Issue: 6

Pages: 10361-10380

Print publication date: 01/06/2014

Online publication date: 12/06/2014

Acceptance date: 30/05/2014

Date deposited: 05/09/2014

ISSN (electronic): 1424-8220

Publisher: MDPI AG


DOI: 10.3390/s140610361


Altmetrics provided by Altmetric


Funder referenceFunder name
51107080National Natural Science Foundation of China