Acoustic emission source positioning research of 3D braided composite material based on the wavelet network

The acoustic emission detection technology is used to position the acoustic emission source of 3D braided composite material. Through comprehensive utilization of the characteristic parameters of acoustic emission signals, the wavelet neural network (WNN) is used to conduct damage positioning and co...

Full description

Saved in:
Bibliographic Details
Published in:Functional Materials
Date:2016
Main Authors: Su Hua, Zhang Tianyuan, Zhang Ning
Format: Article
Language:English
Published: НТК «Інститут монокристалів» НАН України 2016
Subjects:
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/120633
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:Acoustic emission source positioning research of 3D braided composite material based on the wavelet network / Su Hua, Zhang Tianyuan, Zhang Ning // Functional Materials. — 2016. — Т. 23, № 2. — С. 331-336. — Бібліогр.: 8 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1860249858050359296
author Su Hua
Zhang Tianyuan
Zhang Ning
author_facet Su Hua
Zhang Tianyuan
Zhang Ning
citation_txt Acoustic emission source positioning research of 3D braided composite material based on the wavelet network / Su Hua, Zhang Tianyuan, Zhang Ning // Functional Materials. — 2016. — Т. 23, № 2. — С. 331-336. — Бібліогр.: 8 назв. — англ.
collection DSpace DC
container_title Functional Materials
description The acoustic emission detection technology is used to position the acoustic emission source of 3D braided composite material. Through comprehensive utilization of the characteristic parameters of acoustic emission signals, the wavelet neural network (WNN) is used to conduct damage positioning and computation, and by combining the shuffled frog leaping algorithm (SFLA), it can improve the convergence performance. Through experiment comparison with traditional positioning computation method, after optimization with the frog leaping algorithm, the wavelet network acoustic emission source positioning method can effectively improve the precision of damage positioning.
first_indexed 2025-12-07T18:41:48Z
format Article
fulltext Functional materials, 23, 2, 2016 331 ISSN 1027-5495. Functional Materials, 23, No.2 (2016), p. 331-336 doi:http://dx.doi.org/10.15407/fm23.02.331 © 2016 — STC “Institute for Single Crystals” Acoustic emission source positioning research of 3D braided composite material based on the wavelet network Su Hua, Zhang Ning, Zhang Tianyuan Tianjin Polytechnic University, Binshui Road 399, Xiqing District, Tianjin, 300387, P.R. China The acoustic emission detection technology is used to position the acoustic emission source of 3D braided composite material. Through comprehensive utilization of the characteristic param- eters of acoustic emission signals, the wavelet neural network (WNN) is used to conduct damage positioning and computation, and by combining the shuffled frog leaping algorithm (SFLA), it can improve the convergence performance. Through experiment comparison with traditional positioning computation method, after optimization with the frog leaping algorithm, the wavelet network acoustic emission source positioning method can effectively improve the precision of damage positioning. Keywords: Three-dimensional (3D) Braided Composites, shuffled frog leaping algorithm (SFLA),Wavelet Neural Network (WNN), Acoustic Emission Source Positioning Технология детектирования акустической эмиссии используется для определения по- ложения источника акустической эмиссии трехмерных композитных материалов с упроч- няющей оплеткой. Путем всестороннего использования характеристических параметров сигналов акустической эмиссии, вейвлет-нейросети используются для определения и пози- ционирования дефекта, а комбинируя с алгоритмом неупорядоченных прыжков лягушки, можно улучшить сходимость. Путем сравнения экспериментов с традиционным методом расчета позиционирования, после оптимизации алгоритмом прыгающей лягушки, метод вейвлет-нейросетей для позиционирования источника акустической эмиссии может эффек- тивно повысить точность определения положения дефекта. Дослідження позиціонування джерела акустичної емісії тривимірних композитних матеріалів зі зміцнювальним оплетенням на основі вейвлет-мереж. Су Хуа, Чжан Нін, Чжан Тяньюань Технологію детектування акустичної емісії використано для визначення положення джерела акустичної емісії тривимірних композитних матеріалів зі зміцнювальним оплетен- ням. Шляхом всебічного використання характеристичних параметрів сигналів акустичної емісії, вейвлет-нейромережі використовують для визначення та позиціонування дефектів, а завдяки поєднанню з алгоритмом невпорядкованих стрибків жаби можна покращити сход- ження. Шляхом порівняння експериментів з традиційним методом розрахунку позиціону- вання, після оптимізації алгоритмом стрибків жаби, метод вейвлет-нейромереж для пози- ціонування джерела акустичної емісії може ефективно підвищувати точність визначення положення дефектів.. 1. ������������������������ The acoustic emission source positioning is an important work of acoustic emission detec- tion technique, which is an important index to evaluate the acoustic emission detection, and its accuracy reflects whether the position ob- tained through acoustic emission detection is consistent with the actual position where the defect is found. How to increase the position- 332 Functional materials, 23, 2, 2016 Su Hua et al. / Acoustic emission source positioning research ... ing accuracy of acoustic source and maximally reduce missed or false positioning is an impor- tant task during acoustic emission source po- sitioning. Due to different aeolotropic sound velocity, the application of tradition geometri- cal time-difference positioning method for aeo- lotropic composite material is restricted. In re- cent years, in accordance with the aeolotropy of composite material, several acoustic emission source positioning methods based on geometri- cal time-difference positioning method have been developed. For instance, after fitting aeo- lotropic wave velocity, it can be used as known parameter, and added to the classic geometri- cal positioning method and virtual wave front method [1,2,3]. The acoustic emission signal from composite material during loading can include all infor- mation of composite material damage, such as the location of damage source, damage model and degree. It has broad application prospects to use the acoustic emission technology to de- tect the structure of composite material. This paper proposes an acoustic emission detection method of optimized wavelet neural network (WNN) based on the shuffled frog leap- ing algorithm (SFLA). By combining acoustic emission and WNN, this paper conducts theo- retic analysis and specific experiment research on the acoustic emission source positioning of 3D braided composite material. 2. WNN based on shuffled frog leaping algorithm 2.1 Wavelet network The wavelet neural network (WNN) is the product by combining the wavelet analysis theory and neural network theory. The wavelet transform (WT) conducts multi-scale analysis of signal through scale expansion and translation, which can effectively extract local information of signal; the neural network has the character- istics of self-learning, self-adaption and fault tolerance, which is also a common function ap- proximator. The elements and whole structure of WNN are determined in accordance with the wavelet analysis theory, which can avoid the blindness of BP neural network in structural design. In addition, it has stronger learning ability and higher accuracy, and for the same learning task, the WNN has simpler structure and faster convergence rate. [7,8] This paper chooses a compact WNN, and its structure is as shown in Fig.1. Dimension design for input and output. The dimension of input and output should be de- termined in accordance with the problem that needs to be solved and data expression method. In order to conduct accurate positioning of the source of acoustic emission signal, in addition to obtaining traditional time of arrival, we also need to comprehensively consider the sound wave information such as energy, amplitude and counts to conduct positioning training in a clearer and more comprehensive way. Because all the input signals are simulation signals, which requires sampling before training, the di- mension of input layer should be determined in accordance with the sampling point number of waveform, which should be finally determined by coordinating the positioning accuracy. The dimension of output layer should also be deter- mined in accordance with actual requirement. The dimension of output layer should also be determined with actual requirement. This pa- per mainly studies the source positioning of acoustic emission signal, so the node number of input layer is set as 8, and the nerve cell num- ber at output layer is set as 2, which are the abscissa and ordinate respectively. Excitation function. By combining the char- acteristics of damage acoustic emission signal of 3D braided composite material, this paper chooses the Morlet wavelet as the network ex- citation function: ψ ( ) cos( . ) exp( . )x x x= -1 75 0 5 2 (1) Through translation and expansion of ψ(x), we can obtain the j(j = 1,2,...,J)th wavelet function on the hidden layer of WNN: ψ ψj j j j j jx a x b a a b( ) ,= -æ è ççççç ö ø ÷÷÷÷÷ Î Â1 (2) Hidden layer and node selection. The node number on hidden layer is very important to the performance of whole network. If the number is too low, the network won’t obtain adequate Fig. 1. Three-layers network structure Functional materials, 23, 2, 2016 333 Su Hua et al. / Acoustic emission source positioning research ... information from the sample, which cannot suf- ficiently reflect the internal rule of sample, and it will further reduce the network’s generaliza- tion ability; if the number is too high, it will increase the network’s learning time, which will cause the decrease in convergence rate. At present, there is no explicit analytic expres- sion that can be used to select node number of hidden layer, which is generally determined by combining the actual problem through multiple experiments. By combining the actual situa- tion, in accordance with the error precision re- quirement, this paper sets the node number of hidden layer as 17. 2.2 SFLA The shuffled frog leaping algorithm (SFLA) is a population-based cooperative search meth- od inspired by biological emulation in the na- ture[6]. Its basic idea is: F frogs form the ini- tial frog population, in which, the solution of ith frog in S dimension solution space can be expressed as X x x xi s= ( , ,... ,)1 2 . Then, sort individual frogs in the initial frog population in descending order in accordance with the adaptive value, and find the optimal solution Px for the initial frog population. Later, divide the frog population into m sub-populations, and each sub-population contains n frogs, which sat- isfy the relation F = m×n. When the first frog enters the first sub-population, the second frog enters the second sub-population, the mth frog enters the mth sub-population, and the (m + 1)th frog enters the first sub-population, until all frogs have entered the specified sub-popula- tion. In Formula (3), j refers to the jth frog, while Pj refers to the probability of the jth frog being chosen. Then, we can find the best solution Pb and worst solution Pw in the sub-population; in accordance with Formulas (4) and (5), we can conduct local deep search to each sub-popula- tion [7,8]. P n j n n j nj = + - + =2 1 1 1 2( ) /[ ( )], , ,..., (3) S Rand P Pb w= ´ -() ( ) (4) P P S S S Snw w= + - £ £,( )max max (5) In Formula (4), Rand() refers to a random number between 0 and 1; S refers to the frog’s leap step size, which is a reasonable difference value; Smax: maximum step size of each move- ment made by single frog; Pnw refers to the up- dated Pw. If Pw is in a feasible solution space, calculate corresponding fitness of Pnw. If the corresponding fitness of Pnw is less optimal than the corresponding fitness of Pw replace Pb in Formula (4) with Px to update Pw; if there still is no progress, randomly generate a new frog to replace Pw. Otherwise, repeat the update pro- cess until it reaches the preset local iterations LS. After completing deep search of various sub-populations, remix and sort the frogs in the frog population; then, divide the frog popu- lation into various sub-populations to continue local search, until it satisfies any preset condi- tion to stop the algorithm. In this paper, the SFLA is used to optimize the WNN, the initial operation parameters are set as: F: the population size is 100, which refers to the total quantity of individuals contained in the population. m: the sub-population size is 20. Smax: the maximum step size of each move- ment made by single frog is 20. LS: the local evolution times are 10. SF: the glocal evolution times are 400. 3. Experimental 3.1 Experimental sample The TORAYCCA@ T300 carbon fiber (12K) is used as the reinforced fiber for the plate spec- imen of 3D braided composite material. The braided structure of fabricated part is four-step 1×1 3D 5-direction braided structure, and the basis material is TDE86# epoxy resin. The res- in transfer molding (RTM) process is used for curing molding, and the sample has a thickness of (5±0.1)mm. In the experiment, 6 samples of the same size (35cm×20cm×5mm) are chosen and divided into Group A and Group B, with 3 samples in each group; Group A is used in the lead-break experiment, while Group B is used in the damage positioning experiment. 3.2 Four-probe array plane positioning method This test adopts the four-probe array plane positioning method to position the acoustic emission source, as shown in Fig.2. Arrange the four probes into a rhombus on the surface of test piece, the time difference of probes S1 and S3 in receiving the acoustic emis- sion signals is ∆tx, while the time difference of probes S2 and S4 in receiving the acoustic emis- sion signals is ∆ty. The distance between S1 and S3 is a, while the distance between S2 and S4 is b, and the wave velocity is V. Therefore, the AE source is located at the intersection point Q(X,Y) between two curves, and the coordinate expressions are as shown in Formulas (6) and (7). X t V a t V X a Yx x= D D + - æ è ççç ö ø ÷÷÷÷ + æ è çççççç ö ø ÷÷÷÷÷÷÷2 2 2 2 2 (6) 334 Functional materials, 23, 2, 2016 Su Hua et al. / Acoustic emission source positioning research ... Y t V b t V Y a Xy y= D D + - æ è ççç ö ø ÷÷÷÷ + æ è çççççç ö ø ÷÷÷÷÷÷÷2 2 2 2 2 (7) 3.3 Result and analysis of simulative lead-break experiment Conduct four-point positioning experiment to the plate test piece of 3D four-direction braided composite material in Group A. Fig.3 shows the diagram of four-direction plane lead-break experiment, and the four sen- sors are arranged into a rhombus. The standard lead-break experiment is used to simulate the source which generates the acoustic emission signals, and take the average value of three re- sponses. In the experiment, with the rhombus center as the origin, and the coordinates of four sensors are (0,15); (0,–15); (–30,0) and (30,0) respectively. Select four test points with the co- ordinates of (15,10), (–15,–10), (–5,5) and (5,–5) respectively, conduct 3 experiments at each test point, and calculate the average value. Use traditional mathematical morphology to conduct filtering processing of the experiment data obtained above; then, use the correlation analysis method to extract critical values of time difference; conduct computation to obtain the coordinate location of acoustic emission sig- nals. In the meantime, for the convenience of comparison, the positioning analysis method that combines the WNN and SFLA is adopted, which integrates multiple other parameters, including the time difference parameter, into the positioning algorithm to obtain better posi- tioning result. Because the experiment results of test pieces in Group A are close, Test piece 1is used as the example to analyze the experi- ment result. Table 1 shows the experiment data of Test piece 1. n accordance with the ex- periment results, we can see that when single parameter time difference is used to position the test point, even after mathematic morpho- logical filtering and correlation analysis, its positioning accuracy is still low. The absolute average error of its abscissa is 2.1 cm, while Fig. 2. Four-probe array AE source positioning method Fig.3 Sensor layout on the plate test piece of 3D four-direction braided composite material dur- ing lead-break experiment Table 1 Four-point positioning result comparison for Test piece 1 in Group A Four-point positioning comparison of 3D braided composite material Test point (cm) WNN + SFLA positioning method Correlation analysis method Positioning value (cm) Error (cm) Positioning value (cm) Error (cm) No. x y x y x y x y x y 1 15 10 16.5 11.6 1.5 1.6 12.7 12.1 2.3 2.1 2 15 10 13.4 11.3 1.6 1.3 13.1 12.4 1.9 2.4 3 15 10 16.4 11.5 1.4 1.5 12.5 12.5 2.5 2.5 4 –15 –10 –13.7 –8.7 1.3 1.3 –13.5 –8.7 1.5 1.3 5 –15 –10 –16.4 –11.6 1.4 1.6 –12.4 –7.9 2.6 2.1 6 –15 –10 –13.5 –11.5 1.5 1.5 –12.6 –7.6 2.4 2.4 7 5 –5 6.3 –3.6 1.3 1.4 3.2 –6.8 1.8 1.8 8 5 –5 6.8 –3.5 1.8 1.5 2.9 –7.1 2.1 2.1 9 5 –5 3.8 –3.8 1.2 1.2 2.8 –6.9 2.2 1.9 10 –5 5 -3.4 6.5 1.6 1.5 –3.2 6.6 1.8 1.6 11 –5 5 -6.3 6.4 1.3 1.4 –2.9 7.1 2.1 2.1 12 –5 5 -3.5 6.5 1.5 1.5 –3.0 7.2 2 2.2 Functional materials, 23, 2, 2016 335 Su Hua et al. / Acoustic emission source positioning research ... absolute average error of its ordinate is 2.04, which is caused by the anisotropy of composite material in acoustic wave propagation velocity. In the meantime, the unique characteristics of braided angle also affect the measured value of velocity of the acoustic emission signals, while this also directly affects the measurement pre- cision of time difference, which causes big error. When the WNN and SFLA are combined to pro- cess data and conduct positioning analysis, the absolute average error of its abscissa is 1.45cm, while absolute average error of its ordinate is 1.44. We can see that the positioning error has declined by 5mm, and the positioning accuracy has been significantly improved. 3.4 Result and analysis of damage posi- tioning experiment of 3D braided compos- ite material Test pieces in Group B is chosen for the ex- periment, and for the convenience to compare the test results, we artificially created a dotted surface damage on each test piece in advance, as shown in Fig.4. We place the sensors on the coordinates of (0,15); (0,–15); (–30,0) and (30,0) respectively. In this paper, Test piece 1 in Group B is chosen for analysis. After causing dotted damage at coordinate (–13, 6) on the test piece surface, stress is imposed on the test piece for 8 consecutive times through the impact test machine, and the acoustic emission signals are measured and recorded. By adopting the posi- tioning method that combines WNN and SFLA, the coordinate position of damage is calculated, as shown in Table 2. In accordance with the acoustic emission positioning experiment of 3D braided com- posite material, we can see that based on the traditional positioning method, by adding morphological filtering, correlation processing and related neural network technology based on SFLA, we can filter the mechanical noise, conduct multi-factor determination of relative time difference and study the multi-parameter self-learning positioning method, and great re- sults have been achieved. During the damage positioning, the absolute average error of its abscissa is 1.45cm, while absolute average error of its ordinate is 1.44, which is consistent with the error range of Group A during lead-break experiment. The damage assessment points present even distribution with the damage point as the center, as shown in Fig.5. In Fig.5, the round spot represents the actual damage position, while the asterisk represents the neu- ral network and frog leaping tagged position. In accordance with Fig.5, we can see that by adopting the positioning method that combines WNN and SFLA, the positioning accuracy has been significantly improved. 4. Conclusion During acoustic emission positioning re- search on the damage of 3D braided compos- ite material, through experiemnt, it can verify that: by adopting the positioning method that combines WNN and SFLA, the positioning ac- curacy has been significantly improved. Com- pared to the traditional positioning analysis Table 2 Actual damage positioning results of Test piece1 in Group B Times Actual co- ordinates (cm) Computed coordinates (cm) Error (cm) x y x y x y 1 –13 6 –11.5 7.7 1.5 1.7 2 –13 6 –11.4 7.3 1.6 1.3 3 –13 6 –14.8 4.7 1.8 1.3 4 –13 6 –14.4 4.4 1.4 1.6 5 –13 6 –11.2 7.5 1.8 1.5 6 –13 6 –11.6 4.9 1.4 1.1 7 –13 6 –14.4 7.6 1.4 1.6 8 –13 6 –14.5 7.4 1.5 1.4 Fig. 4. Damage position of 3D braided composite material Fig.5 Actual damage position comparison of 3D braided composite material 336 Functional materials, 23, 2, 2016 Su Hua et al. / Acoustic emission source positioning research ... method, the positioning error has declined by 5mm on average. Due to the impact of various factors of 3D braided composite material on the extraction of acoustic emission signals, such as its braided angle, fiber saturated level and resin infiltra- tion/curing, it cannot realize accurate position- ing of damage. Refe�e��es 1. Wan ZhenKai, J. Text. Res., 28, 53, 2007 2. Gu Haibei, Liu Wugang, Sun Fei, Zhang Kai, Missiles and Space Vehicles of Chine, 1, 49, 2012. 3. Liu Zhidong, Pang Baojun, Tang Qi., Piezoelec- trics & Acoustooptics, 32, 493, 2010. 4. Huang XiangSheng, Xiao HanBin, Port Engi- neer.Techn., 3, 13, 2008 5. Cai QiYing, Lin JiangHua, J. Vibr. Shock., 21(3), 11, 2002. 6. Feng H, Liang R Y, Li D X, et al.First IEEE In- ternational Conference on Information Science and Engineering. IEEE Computer Society, p. 3600,2009. 7. Rao R M, Lakshmi. J, Comp. Mater., 46 (24): 3031, 2012. 8. Bijami E, Shahriari-Kahkeshi M, Zekri M., In- ternational Conference on Intelligent Informa- tion Technology Application. 2010.
id nasplib_isofts_kiev_ua-123456789-120633
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1027-5495
language English
last_indexed 2025-12-07T18:41:48Z
publishDate 2016
publisher НТК «Інститут монокристалів» НАН України
record_format dspace
spelling Su Hua
Zhang Tianyuan
Zhang Ning
2017-06-12T14:29:39Z
2017-06-12T14:29:39Z
2016
Acoustic emission source positioning research of 3D braided composite material based on the wavelet network / Su Hua, Zhang Tianyuan, Zhang Ning // Functional Materials. — 2016. — Т. 23, № 2. — С. 331-336. — Бібліогр.: 8 назв. — англ.
1027-5495
DOI: dx.doi.org/10.15407/fm23.02.331
https://nasplib.isofts.kiev.ua/handle/123456789/120633
The acoustic emission detection technology is used to position the acoustic emission source of 3D braided composite material. Through comprehensive utilization of the characteristic parameters of acoustic emission signals, the wavelet neural network (WNN) is used to conduct damage positioning and computation, and by combining the shuffled frog leaping algorithm (SFLA), it can improve the convergence performance. Through experiment comparison with traditional positioning computation method, after optimization with the frog leaping algorithm, the wavelet network acoustic emission source positioning method can effectively improve the precision of damage positioning.
en
НТК «Інститут монокристалів» НАН України
Functional Materials
Technology
Acoustic emission source positioning research of 3D braided composite material based on the wavelet network
Article
published earlier
spellingShingle Acoustic emission source positioning research of 3D braided composite material based on the wavelet network
Su Hua
Zhang Tianyuan
Zhang Ning
Technology
title Acoustic emission source positioning research of 3D braided composite material based on the wavelet network
title_full Acoustic emission source positioning research of 3D braided composite material based on the wavelet network
title_fullStr Acoustic emission source positioning research of 3D braided composite material based on the wavelet network
title_full_unstemmed Acoustic emission source positioning research of 3D braided composite material based on the wavelet network
title_short Acoustic emission source positioning research of 3D braided composite material based on the wavelet network
title_sort acoustic emission source positioning research of 3d braided composite material based on the wavelet network
topic Technology
topic_facet Technology
url https://nasplib.isofts.kiev.ua/handle/123456789/120633
work_keys_str_mv AT suhua acousticemissionsourcepositioningresearchof3dbraidedcompositematerialbasedonthewaveletnetwork
AT zhangtianyuan acousticemissionsourcepositioningresearchof3dbraidedcompositematerialbasedonthewaveletnetwork
AT zhangning acousticemissionsourcepositioningresearchof3dbraidedcompositematerialbasedonthewaveletnetwork