Dissertation•
신경회로망과 웨이블릿 변환을 이용한 스마트 복합적층판의 충격 모니터링 연구 = Impact monitoring of smart composite laminates using neural networks and wavelet analysis
01 Jan 2001-
TL;DR: In this paper, the authors discuss the process for impact location detection in which the generated acoustic signals are detected by PZT using the improved neural network paradigms, and apply the Levenberg-Marquardt algorithm and the generalization methods to improve the accuracy and reliability of a neural network based impact identification method.
Abstract: Low-velocity impact damage is a major concern in the design of structures made of advanced laminated composites, because such damage is mostly hidden inside the laminates and cannot be detected by visual inspection. It is necessary to develop the impact monitoring techniques providing on-line diagnostics of smart composite structures susceptible to impacts. In this paper, we discuss the process for impact location detection in which the generated acoustic signals are detected by PZT using the improved neural network paradigms. To improve the accuracy and reliability of a neural network based impact identification method, the Levenberg-Marquardt algorithm and the generalization methods were applied. This study concentrates not only on the determination of the location of impacts from sensor data, but also the implementation of time-frequency analysis such as the Wavelet Transform (WT) to measure the characteristic frequencies of acoustic emission waves for the determination of the occurrence and the estimation of impact damage.
Citations
More filters
TL;DR: A comprehensive review on the state of the art of Lamb wave-based damage identification approaches for composite structures, addressing the advances and achievements in these techniques in the past decades, is provided in this paper.
1,350 citations
TL;DR: In this paper, the authors presented a new in situ Structural Health Monitoring (SHM) system able to identify the location of acoustic emission (AE) sources due to low-velocity impacts and to determine the group velocity in complex composite structures with unknown lay-up and thickness.
Abstract: This paper presents a new in situ Structural Health Monitoring (SHM) system able to identify the location of acoustic emission (AE) sources due to low-velocity impacts and to determine the group velocity in complex composite structures with unknown lay-up and thickness. The proposed algorithm is based on the differences of stress waves measured by six piezoelectric sensors surface bonded. The magnitude of the Continuous Wavelet Transform (CWT) squared modulus was employed for the identification of the time of arrivals (TOA) of the flexural Lamb mode ( A 0 ). Then, the coordinates of the impact location and the flexural wave velocity were obtained by solving a set of non-linear equations through a combination of global Line Search and backtracking techniques associated to a local Newton’s iterative method. To validate this algorithm, experimental tests were conducted on two different composite structures, a quasi-isotropic CFRP and a sandwich panel. The results showed that the impact source location and the group speed were predicted with reasonable accuracy (maximum error in estimation of the impact location was approximately 2% for quasi-isotropic CFRP panel and nearly 1% for sandwich plate), requiring little computational time (less than 2 s).
175 citations
TL;DR: In this article, a combination of unconstrained optimization technique associated with a local Newton's iterative method was employed to solve a set of nonlinear equations in order to assess the impact location coordinates and the wave speed.
Abstract: This paper investigates the development of an in situ impact detection monitoring system able to identify in real-time the acoustic emission location. The proposed algorithm is based on the differences of stress waves measured by surface-bonded piezoelectric transducers. A joint time-frequency analysis based on the magnitude of the continuous wavelet transform was used to determine the time of arrival of the wavepackets. A combination of unconstrained optimization technique associated with a local Newton's iterative method was employed to solve a set of nonlinear equations in order to assess the impact location coordinates and the wave speed. With the proposed approach, the drawbacks of a triangulation method in terms of estimating a priori the group velocity and the need to find the best time-frequency technique for the time-of-arrival determination were overcome. Moreover, this algorithm proved to be very robust since it was able to converge from almost any guess point and required little computational time. A comparison between the theoretical and experimental results carried out with piezoelectric film (PVDF) and acoustic emission transducers showed that the impact source location and the wave velocity were predicted with reasonable accuracy. In particular, the maximum error in estimation of the impact location was less than 2% and about 1% for the flexural wave velocity.
133 citations
TL;DR: In this article, a methodology for impact identification on composite stiffened panels using piezoceramic sensors has been presented, where a large number of impacts covering a wide range of energies (corresponding to small and large mass impacts) at various locations of a composite stiffening panel have been simulated using the finite element (FE) method.
Abstract: In this work a methodology for impact identification on composite stiffened panels using piezoceramic sensors has been presented. A large number of impacts covering a wide range of energies (corresponding to small and large mass impacts) at various locations of a composite stiffened panel have been simulated using the finite element (FE) method. To predict the impact location, artificial neural networks have been established using the data generated from FE analyses. A number of sensor signal features have been examined as inputs to the neural network and the effect of noise on the predictions has been investigated. The results of the study show that the trained network is capable of locating impacts with different energies at different locations (e.g. in the bay, over/under the stringer and on the foot of the stringer) in a complicated structure such as a composite stiffened panel.
125 citations
TL;DR: In this paper, the authors used an efficient nonlinear finite element (FE) model of a sensorized composite stiffened panel to reconstruct the impact force history using artificial neural networks (ANNs) and spectral components of sensor data recorded by piezoceramic sensors.
Abstract: In this work, a new methodology is presented for reconstruction of the impact force history using artificial neural networks (ANNs) and spectral components of sensor data recorded by piezoceramic sensors. A large set of data, required for training the ANNs, was generated by using an efficient nonlinear finite element (FE) model of a sensorised composite stiffened panel. Impact experiments were performed on a composite plate equipped with surface-mounted piezoceramic sensors to validate the numerical modelling approach. Using the FE model of the panel, data were generated for impacts that are likely to occur during the life-time of an aircraft, consisting of large mass (e.g. dropping tool) and small mass (e.g. debris) impacts at various locations, i.e. in the bay, on the foot of a stringer and over/under a stringer. Even though the panel undergoes large deformation during impact (nonlinear response), the established networks predict the impact force history and its peak with reasonable accuracy.
110 citations
References
More filters
TL;DR: The Marquardt algorithm for nonlinear least squares is presented and is incorporated into the backpropagation algorithm for training feedforward neural networks and is found to be much more efficient than either of the other techniques when the network contains no more than a few hundred weights.
Abstract: The Marquardt algorithm for nonlinear least squares is presented and is incorporated into the backpropagation algorithm for training feedforward neural networks. The algorithm is tested on several function approximation problems, and is compared with a conjugate gradient algorithm and a variable learning rate algorithm. It is found that the Marquardt algorithm is much more efficient than either of the other techniques when the network contains no more than a few hundred weights. >
6,899 citations
TL;DR: A comprehensive review on the state of the art of Lamb wave-based damage identification approaches for composite structures, addressing the advances and achievements in these techniques in the past decades, is provided in this paper.
1,350 citations
Book•
13 Oct 1983
TL;DR: This is the first comprehensive text on its subject to appear since the 1960s and incorporates classical material with the many significant developments in the field and is the only up-to-date introduction currently available.
Abstract: This is the first comprehensive text on its subject to appear since the 1960s. It incorporates classical material with the many significant developments in the field and is the only up-to-date introduction currently available."Introduction to Random Vibrations "presents a brief review of probability theory, a concise treatment of random variables and random processes (including normal, Poisson, and Markov processes), and a comprehensive exposition of the theory of random vibrations.It contains a number of noteworthy features. Linear systems theory is introduced with a high degree of generality in order to demonstrate its elegance and range of applicability. The response of discrete and continuous linear systems to random excitations is then developed within this framework. The chapter on the response of nonlinear systems represents a unified view of the topic, incorporating some major recent formulations. The discrete-state approach, which has emerged as a powerful technique, is utilized in the treatment of a number of random process properties, among them level crossings, peaks, envelopes, and first-passage times. The Stieltje integral representation of random processes is introduced in order to simplify the presentation of stationary and nonstationary random processes and response statistics.In addition to the opening review of probability and set theory, appendices review relevant topics in Fourier analysis and ordinary differential equations. Both these reviews and exercises included with the chapters will be useful to students using the book as a course text and to practitioners using it as a reference.This book is third in The MIT Press Series in Structural Mechanics, edited by Max Irvine.
650 citations
TL;DR: In this paper, the authors used the wavelet transform to represent all possible types of transients in vibration signals generated by faults in a gearbox and demonstrated the application of the suggested wavelet by a simple computer algorithm.
490 citations
TL;DR: In this paper, the authors presented a new in situ Structural Health Monitoring (SHM) system able to identify the location of acoustic emission (AE) sources due to low-velocity impacts and to determine the group velocity in complex composite structures with unknown lay-up and thickness.
Abstract: This paper presents a new in situ Structural Health Monitoring (SHM) system able to identify the location of acoustic emission (AE) sources due to low-velocity impacts and to determine the group velocity in complex composite structures with unknown lay-up and thickness. The proposed algorithm is based on the differences of stress waves measured by six piezoelectric sensors surface bonded. The magnitude of the Continuous Wavelet Transform (CWT) squared modulus was employed for the identification of the time of arrivals (TOA) of the flexural Lamb mode ( A 0 ). Then, the coordinates of the impact location and the flexural wave velocity were obtained by solving a set of non-linear equations through a combination of global Line Search and backtracking techniques associated to a local Newton’s iterative method. To validate this algorithm, experimental tests were conducted on two different composite structures, a quasi-isotropic CFRP and a sandwich panel. The results showed that the impact source location and the group speed were predicted with reasonable accuracy (maximum error in estimation of the impact location was approximately 2% for quasi-isotropic CFRP panel and nearly 1% for sandwich plate), requiring little computational time (less than 2 s).
175 citations