ADVANCED STOCHASTIC FEM-BASED ARTIFICIAL NEURAL NETWORK FOR CRACK DAMAGE DETECTION

被引:0
作者
Sbarufatti, C. [1 ]
Manes, A. [1 ]
Giglio, M. [1 ]
机构
[1] Politecn Milan, Dipartimento Meccan, I-20156 Milan, Italy
来源
COMPUTATIONAL METHODS FOR COUPLED PROBLEMS IN SCIENCE AND ENGINEERING IV | 2011年
关键词
Structural Health Monitoring; Artificial Neural Network; structural diagnosis; crack; helicopter;
D O I
暂无
中图分类号
T [工业技术];
学科分类号
08 ;
摘要
Structural Health Monitoring (SHM) is nowadays one of the most challenging research fields. As a matter of fact, if from one hand the aerospace industry is trying to extend the duration of life-limited components, from the other hand a deep control is necessary over the structures to guarantee both the machine availability and reliability. In effect, thanks to the advance in the evaluation of the actual structural health by means of a SHM system, it could be possible to set a Condition Based Maintenance (CBM). This approach means substituting a component according to its real structural conditions instead of relying just on the design assumptions. The final aim is to update the scheduled maintenance intervals according to the actual condition of the structures. However this is not an easy task, as it is governed and influenced by many variables, each one characterized by a stochastic distribution. In particular, the key factor is the disposal of detection and monitoring systems as reliable as possible in order to conjugate safety with economics objective. On the basis of this all the machine stops can be optimized in order to exploit the machine availability with the minimum loss of reliability. Thus, the first step for developing such advanced technology would be the disposal of a robust damage detection system, able to recognise, locate and quantify the damage in a certain component. The aim of the present work is to define a methodology that combines the use of Finite Element Models (FEM) with Artificial Neural Networks (ANN) till for crack detection over a typical aerospace structure consisting of a riveted aluminium skin stiffened with some reinforcing elements [2]. Numerical models, in fact, could be used to train ANN. A basic system knowledge would result, upon which to introduce the variability by means of real sensor network data [3], in order to consider the problem from a statistical point of view. Finally, a proposal for the sensor network characterization in terms of Probability of Detection (PoD) and False Alarm (PFA) is also reported.
引用
收藏
页码:1107 / 1119
页数:13
相关论文
共 13 条
[1]  
[Anonymous], NASACR2002211428
[2]  
[Anonymous], 1997, MACHINE LEARNING, MCGRAW-HILL SCIENCE/ENGINEERING/MATH
[3]  
Bishop CM., 1995, NEURAL NETWORKS PATT
[4]   A Bayesian sampling approach to decision fusion using hierarchical models [J].
Chen, B ;
Varshney, PK .
IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2002, 50 (08) :1809-1818
[5]   Crack propagation on helicopter panel: Experimental test and analysis [J].
Giglio, M. ;
Manes, A. .
ENGINEERING FRACTURE MECHANICS, 2008, 75 (3-4) :866-879
[6]   Surface crack detection for Al plate using the surface acoustic waves and neural network identification [J].
Guan, JF ;
Shen, ZH ;
Xu, BQ ;
Lu, J ;
Ni, XW .
LASERS IN MATERIAL PROCESSING AND MANUFACTURING II, 2005, 5629 :466-474
[7]  
Hideki S., 2006, SOURCE APPL COMPOSIT, V13/2
[8]  
Hoballah I. Y., 1989, IEEE T INFORM THEORY, V35
[9]  
Katsikeros C.E., 2009, MECH SYSTEMS SIGNAL, V23/2
[10]   Artificial Neural Network (ANN)-based Crack Identification in Aluminum Plates with Lamb Wave Signals [J].
Lu, Ye ;
Ye, Lin ;
Su, Zhongqing ;
Zhou, Limin ;
Cheng, Li .
JOURNAL OF INTELLIGENT MATERIAL SYSTEMS AND STRUCTURES, 2009, 20 (01) :39-49