Simulated Acoustic Emission Signals Research Articles

Acoustic Emission Signal Denoising of Bridge Structures Using SOM Neural Network Machine Learning

Identification of noise signals is one of the most challenging problems in health monitoring of a bridge structure using acoustic emission (AE) monitoring and identification technology. Hardware filtering technology and spatial identification technologies are the most common methods used in identifying signals from the defects of the bridge, which have great limitations due to the presence of environmental noise. Therefore, this paper focuses on the AE noise signal from a bridge in an operation state and other specific loading states, which is diagnosed in the hardware filtering technology, spatial identification, and self-organizing map (SOM) neural network, to obtain the new noise recognition methods. It is found that the first two methods can indeed filter the noise signal, but the filtering rate can only reach about 50%, and can barely filter strong noise signals. The SOM neural network had strong self-recognition ability. The classification accuracy of simulated AE signals is 90% and 100%, respectively. The trained network was used to test 183 sample signals, and the defect signal detection accuracy reached 76% and 78.8%; therefore, the noise signal filtering effect is significantly improved.

Signal Identification of Wire Breaking in Bridge Cables Based on Machine Learning

With the booming development of bridge construction, bridge operation and maintenance have always been major issues to ensure the safety of the community. Affected by the long-term service of bridges and natural factors, the safety and durability of cables can be threatened. Cables are critical stress-bearing elements of large bridges such as cable-stayed bridges. Realizing the health monitoring of bridge cables is the key to ensuring the normal operation of bridges. Acoustic emission (AE) is a dynamic nondestructive testing method that is increasingly used in the local monitoring of bridge cables. In this paper, a testbed is described for generating the acoustic emission signals for signal identification testing with machine learning (ML) models. Owing to the limited number of measured signals being available, an algorithm is proposed to simulate acoustic emission signals for model training. A multi-angle feature extraction method is proposed to extract the acoustic emission signals and construct a comprehensive feature vector to characterize the acoustic emission signals. Seven ML models are trained with the simulated acoustic emission signals. Long short-term memory (LSTM) has been specially applied for deep learning demonstration which requires a large amount of training data. As all machine learning models (including LSTM) provide desired performance, it shows that the proposed approach of simulating acoustic emission signals can be effective.

Acoustic emission source modeling using a data-driven approach

The next generation of acoustics-based non-destructive evaluation for structural health monitoring applications will depend, among other reasons, on the capability to effectively characterize the transient stress wave effects related to acoustic emission (AE) generated due to activation of failure mechanisms in materials and structures. In this context, the forward problem of simulating AE is addressed herein by a combination of experimental, analytical and computational methods, which are used to form a data-driven finite element (FE) model for AE generation and associated transient elastic wave propagation. Acoustic emission is viewed for this purpose as part of the dynamic process of energy release caused by crack initiation. To this aim, full field experimental data obtained from crack initiation monitored by digital image correlation is used to construct a traction-separation law and to define damage initiation parameters. Subsequently, 3D FE simulations based on this law are performed using both a cohesive and an extended finite element modeling approach. To create a realistic computational AE source model, the transition between static and dynamic responses is evaluated. Numerically simulated AE signals from the dynamic response due to the onset of crack growth are analyzed in the context of the inverse problem of source identification and demonstrate the effects of material and geometry in crack-induced wave propagation.

Simulation and Experimental Investigation on the AE Tomography to Improve AE Source Location in the Concrete Structure

Acoustic emission (AE) tomography, which is based on the time-travel tomography with AE events as its signal sources, is a new visualization tool for inspecting and locating the internal damages in the structures. In this paper, AE tomography is applied to examine a man-made damage in a typical heterogeneous concrete structure to validate its effectiveness. Firstly, the finite element (ABAQUS/Explicit) simulation model of the concrete structure with one damaged circle in its center is built, and the simulated AE signals are obtained to establish the AE tomography. The results show that the damaged circle in the created model can be visualized clearly with the AE tomography in its original location. Secondly, the concrete specimen based on the FE model is fabricated, and the pencil lead break (PLB) signal is taken as the exciting source for AE tomography. It is shown that the experimental results have good consistency with the FE simulation results, which also verifies the feasibility of the finite element model for AE tomography. Finally, the damage source location based on AE tomography is compared with the traditional time of arrival (TOA) location method, and the better location accuracy is obtained with the AE tomography. The research results indicate that AE tomography has great potential in the application of structure damage detection.

Particle filter based noise removal method for acoustic emission signals

This paper discusses the application of a statistical noise removal technique – Rao–Blackwellised particle filter (RBPF) for signal to noise ratio (SNR) enhancement of acoustic emission (AE) signals. RBPF is a recursive Bayesian method for dynamic system state estimation. Compared with other signal filtering methods, RBPF offers the advantage of broad-band signal cleansing by directly modeling the internal dynamics of the concerned physical system and statistical characteristics of the signal noise. In doing so, signal filtering can be related to the dynamic characteristics of the underlying physical system, rather than a purely mathematical operation. RBPF also outperforms a few other statistical signal filtering methods such as Kalman filter, with the ability of handling nonlinear system and non-Gaussian noise problems. Another feature of RBPF is its ability to allow real-time on-board signal processing. In this paper, moment tensor analysis was performed first to generate simulated baseline AE signals. The simulated AE signal was subsequently superimposed with noise to demonstrate the effectiveness of the RBPF method in filtering noise in the AE signals. The results show that the performance of the RBPF in SNR enhancement is very promising. Finally, RBPF is also applied to real AE data obtained from experimental tests and apparent improvement to the SNR in AE feature study is observed.

Feature extraction of AE characteristics in offshore structure model using Hilbert–Huang transform

This paper addresses an application of recently developed signal processing tool based on the Hilbert–Huang transform (HHT) to characterize the acoustic emission (AE) signals released from the offshore structure model. The AE signals from the cracks in the welded steel nodes of the offshore structure model are collected during the tensile testing in water and the simulated AE signals are also collected in the offshore structure model. Instantaneous frequency and energy features based on local properties of the AE signals are then extracted using the Hilbert–Huang transform method. In order to demonstrate the advantage of the AE feature extraction techniques based on Hilbert–Huang transform, the conventional AE analysis is also provided side-by-side for comparison. The results verify that the method based on HHT better characterizes the AE signals than the classical AE techniques. It can be concluded that the AE signal analysis based on HHT is an effective tool to extract the features and this opens perspectives for crack recognition in offshore structures.

Simulation of Acoustic Emission in Planar Carbon Fiber Reinforced Plastic Specimens

The simulation of acoustic emission waveforms resulting from failure during mechanical loading of carbon fiber reinforced plastic structures is investigated using a finite element simulation approach. For this investigation we focus on the dominant failure mechanisms in fiber reinforced structures consisting of matrix cracking, fiber breakage and fiber-matrix interface failure. To simulate the failure process accurately, we present a new acoustic emission source model that is based on the microscopic source geometry and micromechanical properties of fiber and resin. We demonstrate that based on this microscopic source model these failure mechanisms result in excitation of macroscopic plate waves. The propagation of these plate waves is described using a macroscopic three-dimensional model geometry which includes contributions of reflections from the specimen boundaries. We further present a model of the acoustic emission sensors used in experiments to simulate the influence of aperture effects. To enhance the understanding of correlation between macroscopically detectable acoustic emission signals and microscopic failure mechanisms we simulate the response to different source excitation times, crack surface displacements and displacement directions. The results obtained show good agreement with fundamental assumptions about the crack process reported by various other authors. The simulated acoustic emission signals obtained are compared to experimentally measured waveforms during four-point bending experiments of carbon fiber reinforced plastic structures. The simulated signals of fiber-breakage, matrix-cracking and fiber-matrix interface failure show systematic agreement with the respective experimental signals.

Stabilized fiber-optic extrinsic Fabry–Perot sensor system for acoustic emission measurement

A fiber-optic Fabry–Perot (F–P) acoustic emission (AE) sensor system based on an improved double wavelength stabilization technique is described. Without stabilization, the sensor system drifts out of quadrature due to the presence of low-frequency dynamic strains. The stabilization is achieved by using a dense wavelength division multiplexer (DWDM) to generate two quadrature phase-shifted output signals. An optimum model of double wavelength stabilization is established. The filtering wavelengths of DWDM are calculated using an optimization design method. The performance of the developed sensor system was verified by two preliminary tests. One was a test on the stabilization of operating point of fiber-optic F–P sensor, and other was the detection of simulated AE signals generated by the impact and pencil lead breaking. The test results demonstrate that simulated AE signals are successfully detected using this stabilized sensor system, which solves the fade-out problem.

High-temperature acoustic emission tests using lithium niobate piezocomposite transducers

High-temperature-resistant piezoelectric sensors have been used as acoustic emission sensors over a temperature range from room temperature up to 400°C. The sensors were made from lithium niobate 1-3 piezocomposites, using the dice and fill method, with high-temperature-resistant cement as the passive material. Good results in detecting simulated acoustic emission signals were obtained at high temperatures, with little change from the same measurements at room temperature. Of the devices tested, a 4 mm-thick piezocomposite with 45% volume fraction of y/36°-cut lithium niobate gave the best results. In addition, the possibility to use these devices to locate an acoustic emission source was verified.

Linear Location of Acoustic Emission Sources with a Single Channel Distributed Sensor

In this paper, distributed sensors are evaluated for structural health monitoring. A distributed sensor is formed by electrically connecting a number of acoustic emission sensor nodes in series. The individual nodes are placed at strategic locations on the structure. This arrangement allows the detection of acoustic emissions occurring at multiple regions of the structure using a single monitoring channel, and hence significantly reduces the cost, complexity, and weight of on-board instrumentation. The performance of the distributed sensor is compared against that of conventional single node sensors on a fiberglass bar and on a graphite epoxy sandwich panel using simulated acoustic emission signals. These comparisons indicate that a single distributed sensor can monitor large areas of the structure without degrading the signal amplitude or the frequency bandwidth. A new analysis procedure to determine the location of the acoustic emission source from this single channel data is also introduced. Experiments performed on a steel bar indicate that reasonable accuracy can be achieved in locating the acoustic emission sources along a line segment just using the single channel data from the distributed sensor.

Non-contact measurement of acoustic emission in materials by laser interferometry

The non-contact measuring system of acoustic emission (AE) by laser interferometry was developed to detect AE signals and analyze microfracture quantitatively during materials testing. The capability of this system was estimated by comparison between simulated AE signals due to glass capillary breaking and calculations using the finite element method. The systemcould measure AE signals during practical tensile tests of carbon fiber reinforced plastics. This technique was also applied to the thermal cycle test of ceramic/metal coatings, and AE signals during cooling were successfully detected and analyzed by a deconvolution method to evaluate quantitatively the microfracture process.

Transfer functions for acoustic emission transducers using laser interferometry

A high-fidelity, broadband laser interferometer is used to develop a transfer function for a piezoelectric, acoustic emission transducer. The transfer function, which converts the relative transducer output to an absolute measurement of surface velocity, uses a methodology that accounts for the transducer’s frequency bias. This technique treats the piezoelectric transducer as a corruptive channel and develops a filter that models this bias on a measured acoustic emission signal. This filter, which is developed with a signal processing software package, is based on comparisons between signals measured with a laser interferometer and those made with a piezoelectric transducer. The accuracy of this filter is verified and it is used to remove the transducer-related features from simulated acoustic emission signals.

Extensional and Flexural Waves in a Thin-Walled Graphite/Epoxy Tube*

Simulated acoustic emission signals were induced in a thin-walled graphite/epoxy tube by means of lead breaks (Hsu-Neilsen source). The tube is of similar material and layup to be used by NASA in fabricating the struts of Space Station Freedom. The resulting waveforms were detected by broad band ultrasonic transducers and digitized. Measurements of the velocities of the extensional and flexural modes were made for prop agation directions along the tube axis (0 degrees), around the tube circumference (90 degrees) and at an angle of 45 degrees. These velocities were found to be in agreement with classical plate theory.

Application of an intelligent signal processing system to acoustic emission analysis

This article describes a novel approach for analyzing elastic wave signals to obtain a solution to simple inverse source problems. By using a simulated intelligent system that resembles the structure of a neural network and an acoustic source with its associated field of elastic waves in a plate, a set of pattern vectors is generated. The memory of the system is formed through a learning process in which a systematic series of experiments is presented to the system. In the experiments described here, the system was trained using simulated acoustic emission signals generated by a normal force acting on a thick plate to which information about the source related to its location and characteristics was appended. It is demonstrated, using such a trained system, that one can recover the characteristics of an unknown source from the acoustic signals emitted by it. It is also shown that one can synthesize from the memory the acoustic signals corresponding to an arbitrary source without using any elastodynamic theory.

Monitoring of crack growth and crack closure behavior by a simulated acoustic emission technique.

A new nondestructive method using a simulated acoustic emission (AE) technique is proposed for the monitoring of crack growth and crack closure behavior. Simulated AE signals, which propagated through a crack, were detected and the waveform changes with the crack growth were examined. It became clear that the amplitudes of the first peak of the waveforms have a strong correlation with the crack length and the state of crack closure. The numerical simulation by a finite differential method showed that about ±10mm of crack length can be measured with a pair of AE transducers, and the correlation curve between crack length and amplitude of the simulated AE shows little change with the materials. These results suggest that the simulated AE technique is applicable to the evaluation of both crack propagation and crack closure behavior.

Distortion of acoustic emission signals by a change of gain in the sensing device

A semi-theoretical model of an acoustic emission ( AE) signal is used to predict the effects of changes of amplification and threshold level in the recording device. Predictions are confirmed by experiments on simulated AE signals. The progress of real AE tests on cracks in welds is analysed statistically to predict the effect of the same variations in the sensitivity of the recording device. Once again the predictions are confirmed by experiment and it is shown how to convert an AE/time curve from one set of recording conditions to another. It is concluded that compensation for distortion of AE signals by changes of sensitivity of the recording device presents no great difficulty.