Acoustic Emission testing

Method of acoustic emission signal denoising and signal anal

Update time：2006-11-28 16:52　　　 Viewed：次

Due to the limitation of signal acquisition and processing, the acoustic emission instrument in early time rarely have to transient signal waveform capture and realtime processing ability, therefore that is applied signal analysis is the parameter analysis method. In various application areas, people summarized many experience through years of practice, through the analysis of ae parameters, establish characterization and mapping relationship. Such as the United States ASTM and ASME standards, as well as the Chinese national standard GB/T18182 [1] and so on are the parameters of acoustic emission monitoring objects to nondestructive evaluation and safety evaluation. Prine d. w. [2] for the sake of using acoustic emission evaluation such as the seriousness of the defect in the welding process, to dissect, several kinds of steel structure and material and metallographic method measuring the size of the crack, found that the acoustic emission parameters (total number) and there is close relationship between the crack size. On the fracture mechanics, Dunegan H.L. [3] by using acoustic emission, such as total number of N to expression of stress intensity factor K values in metal materials and the relationship between the crack propagation rate. Rice [4] and Ravindra. H. V [5] also takes advantage of the cutting tool in cutting process of acoustic emission signal energy distribution relationship respectively to predict cutting tool crack extension degree, monitoring the working conditions of cutting tool. Acoustic emission in the study of composite materials, Nanjo [6] and Vaidya [7] are found fiber fracture produce lowrising value of acoustic emission signals (40~60 db), and the intergranular crack produce high amplitude of the acoustic emission signal (65~85 db); Jin Zhou heptyl [8] in the acoustic emission characteristics of carbon/epoxy composite materials in the study found that ringing counts, amplitude, duration, duration and dead load acoustic emission voices than is the difference between composite components, white musk, each stage of injury, injury types, main parameters of mechanical properties.

In parameter analysis method, in order to find the characteristics of acoustic emission source and inherent law, people usually use correlation diagram analysis method, that is the amplitude, duration, energy, time of arrival, RMS voltage value, the number of collisions, the hit rate, external parameters such as correlation analysis between any two variables. From acoustic emission parameter correlation diagram can find the changing rule of the noise signal, to distinguish between different characteristics of signal. As in pressure vessel or pipe, internal medium leakage signal can be considered a continuous signal, thus the container or pipe internal defects compared to AE signals, the signal of the relatively long duration, by some parameters, such as the channel number) for the duration of the connection diagram we can timely found leakage of medium. The MONPAC acoustic emission inspection association makes assessment on the quality of metal pressure vessel AE testing data in the amplitude of the acoustic emission signal counting and associated graph form. Correlation diagram analysis method is the most main parameter analysis method and the commonly used method, people also summarizes some practical experience through years of experiments.

As the demand of engineering practice and developing of acoustic emission techniques and research, according to the need to constantly to define new parameters, the research workers makes the parameter analysis of acoustic emission source inverse problem more deeply based on using the acoustic emission parameters of predecessors' definition. Shiwa put forward to use wave attenuation ratio and evaluation to value the destruction of the thin film layer. In practical application research, we often choose more appropriate composition characteristic parameters of acoustic emission parameters according to actual situation, multiangle multilevel analysis of inverse source problem, to improve the ability of analysis and accuracy. Of course, for some method of parameters, you can also use some modern signal processing methods to find an inner laws and characteristics relevance. Such as gongtian shen application of acoustic emission source metal pressure vessel routine parameters such as energy, duration and amplitude, counting, count and rise time to peak, through the neural network and grey correlation analysis method to distinguish the crack extension success, residual stress and mechanical friction caused by acoustic emission signal. Belchamber in composite materials and civil engineering structures, Murthy and Chichibu application range and duration of conventional parameters, such as pattern recognition, can also distinguish a number of different acoustic emission signal.

Acoustic emission signal waveform analysis

Waveform analysis refers to a signal processing method of getting information of through the analysis of acoustic emission (AE) signals in the time domain waveform or frequency spectrum characteristics. Theoretically, waveform analysis should be able to give any information required, so waveform is also the most accurate method of the expressing characteristics of AE source, and quantitative information of signal can be obtained. So people have realized that in the identification of acoustic emission source waveform analysis and evaluation of the important function of the object under test from the beginning of the acoustic emission technology development, and want to obtain the original waveform signals generated by acoustic emission source through a variety of methods to, but due to short of acquisition and realtime processing hardware and access requirements, and signal processing method is not perfect, so the development of acoustic emission waveform analysis techniques are always restricted. With the rapid development of hardware and software technology, people start to develop full waveform acoustic emission testing instrument, and make use of modern signal processing method for waveform analysis and processing to get more information of acoustic emission source.

Normally, laboratory specimen and engineering components are mainly plate structure, so it becomes more meaningful to study the characteristics of acoustic emission signals in the plate, the difference of the mechanism of different source waveform (and the differences of corresponding frequency spectrum and related characteristics of similarities) and then look for ways to identify. American scholar Gorman and others did a lot of work in terms of the characteristics of acoustic emission waveform composite material plate and put forward the concept of "modal acoustic emission", to distinguish the parameter analysis method people have used to in the past. As shown in figure 3 sheet metal lead is broken (E wave) and the flexible expansion wave produced wave (F).

Figure 3 E and F wave in modal acoustic emission

In the past there were many scholars made deep study in waveform analysis technology. In 50’s, Pekeris and Knopoff , studied the theory of "the epicenter" surface displacement caused by the deep source in the half infinite space (plate) ,the essence of which is to solve Lamb problem of half infinite space under the pulse source. Breckenridge studied surface displacement step produced by load in thick plate, the experiment result of which has become the basis of the transient response on the calibration of AE sensor. Hsu and Sachse make use of convolution integral method and used the numerical calculation, and the step surface displacement, sound effect were or, in turn, by the shape of the sound source surface displacement solution which is used to identify the sound source .
Rongsheng geng makes use of waveform analysis method for evaluating aircraft radome glue failure, and the processing and analysis of the received wave to obtain information on the quality of adhesive by receiving excitation signal propagation in radar cover. Waveform analysis can also be the source of accurate positioning, Ziola etc. through the use of waveform analysis, the method of sound source localization in composite material plate were studied, and obtain the higher positioning accuracy than the method. In short, researchers began to analyze the signal waveform the acoustic emission waveforms, and people usually use the following analysis methods according to the characteristics of the acoustic emission signal.

Classical music and modern spectrum analysis

Spectrum analysis method can be divided into classical spectrum analysis and modern spectrum analysis both of which are the most common analysis methods of acoustic emission signal processing. Based on Fourier transform classic spectral analysis method is simple and convenient, but its resolution is not high, and the spectrum estimation error is bigger. Modern spectrum analysis method use appropriate parameter model to fit signal or mathematical orthogonal signal processing method to separate signal, thus improving the resolution of spectra and statistical stability of spectral estimation. Two methods of spectrum analysis is researching a variety of characteristics of acoustic emission signals in the frequency domain through the acoustic emission signal from time domain to frequency domain, find the eigen information identification of acoustic emission source. Because in theory, acoustic emission signals of different source contain information reflect the features of its nature.

In recent years, people have done a lot of exploration in spectrum analysis and achieved many significant results. Due to acoustic emission technology’s broad application scope, these research work also across all areas. Cawthorne, m. [21]make use of waveform monitoring helicopter rotating system dynamic parts of the spectrum of fatigue, and achieved good effect. Yang [22] found that the acoustic emission signal power spectrum characteristic changes reflect the piston - cylinder liner clearance is different through the automobile engine piston - such as cylinder liner wear process of acoustic emission signal was analyzed, and so using the spectrum index to judge the wear degree of the engine piston, cylinder liner.

In the past few years, research of using the method of spectrum analysis on composite material damage pattern recognition is very much. For example, in the study of four types of injury of epoxy composite material model of the acoustic emission signal characteristic ,De Groot [23] found that frequency spectrum component of different damage model producing acoustic emission signal are not identical and each kind of injury patterns corresponding to a characteristic frequency range; in the research of metal materials’ welding defects, Bentley and Beesleyl [24] analyzed the steel weld defects in acoustic emission signal by using broadband probe and spectrum analysis technology and they found that can identify cracks and slag defect is the most important characteristic frequency component. Liu Shifeng [25] established the corresponding relationship between the typical weld defects acoustic emission source and power spectrum.R.h. ou etc proved that the spectrum of the acoustic emission signal can reflect the characteristics of acoustic emission source, and similar acoustic emission signals of acoustic emission source have similar characteristics of frequency spectrum distribution; Different characteristics of acoustic emission signal of acoustic emission source can be reflected by frequency spectrum distribution information. Therefore, spectrum analysis can reveal the characteristics of acoustic emission source signal and its dynamic characteristics. But the signal spectrum analysis are smooth signal is periodic signal analysis, and spectrum analysis is a kind of global analysis method of ignoring local information changes.

From figure 1 we can see that noise emission signal is a king of nonstationary random signal changes over time, the acoustic emission signal analysis in sometime for detail characteristics is particularly important, therefore spectrum analysis is not an effective method of analyzing the characteristic of the acoustic emission signal, and the resolution is not high and the spectrum estimation of error is bigger. We should find a processing method that is more conform to the characteristics of acoustic emission signal.

The wavelet analysis

Research of signal analysis and processing method is one of the key problems of acoustic emission source identification and evaluation, and is also the main development direction of acoustic emission technology. Neither acute or continuous acoustic emission signal is transient and randomness, belong to nonstationary random signal, and is composed of a series of signal frequency and mode rich. For this type of signal analysis and processing should be able to provide the analysis of time domain and frequency domain, the acoustic emission signal processing method should have the ability of time-frequency analysis.

Wavelet theory is created by mathematicians and engineers from the beginning. From the beginning of the wavelet analysis, it is closely linked with the actual engineering problem, thus it not only has the strict mathematical theory, but also solve many practical engineering problems. At present wavelet analysis is mainly used in image processing, signalnoise separation, speech signal processing, signal feature extraction and fault diagnosis. In the acoustic emission signal processing, it mainly makes use of wavelet analysis the signaltonoise of separation and good timefrequency localization characteristics, to extract the characteristics of acoustic emission signal to find the inherent law, achieve identification of acoustic emission source, understand the purpose of its propagation. Another interesting using, we are interested in some of the frequency of the signal, using the relatively stable signal (no dispersion, the wave velocity must be) time of arrival in different sensors, realize the precise localization of acoustic emission source. Wang. Q combine wavelet transform and correlation analysis of rotor rubbing acoustic emission signal to locate research, obtained the conclusion whose accuracy is higher than that of acoustic emission signal directly related to do high positioning. [27]

Wavelet analysis has brought new vigor for acoustic emission signal processing technology, a large number of literatures have reported the research results wavelet analysis has obtained in the study of acoustic emission signal processing that the other signal processing methods can’t get, which shows that the wavelet analysis is the best approach for the analysis of acoustic emission signals. At the same time, due to acoustic emission testing technology is a practical technology, so it is the study purpose and direction of acoustic emission signal wavelet analysis to apply wavelet analysis to acoustic emission testing projects and solve the practical problems in engineering. But at the same time, every method is not everything and has its scope of application. To truly achieve accurate qualitative, quantitative and orientation for acoustic emission source ,we must also be fully based on the characteristics of acoustic emission signal, combining advantages of various signal processing means, find out the most reasonable analysis method through a large number of engineering inspection.

Identification of neural network

Since 80’ s, the application of artificial intelligence technology ,marked by the technique of artificial neural network (ANN), expert system (ES), has penetrated into all fields, and made many encouraging progress in signal processing, pattern recognition, nonlinear optimization, automatic target recognition, sensing technology, etc. This mainly because they in solving highly nonlinear and uncertain systems the huge potential of reasoning for a variety of difficult to establish accurate mathematical model of the process, such as acoustic emission signal propagation and object recognition brings new hope. The ultimate aim on the acoustic emission signal analysis and processing is to realize the recognition of acoustic emission source, artificial neural network is the main method of pattern recognition, and thus become one of the main content of the acoustic emission signal processing research.

Acoustic emission signal contains audio source mode characteristic information, but because of the communication process, medium characteristics, the influence of the instrument response, from acoustic emission source to the mapping between the acoustic emission signal is nonlinear, it is difficult to for accurate modeling of the uncertain process, which is the advantages of artificial neural network, so the use of artificial neural network for identification of acoustic emission source is a promising way. Artificial neural networks for acoustic emission detection technology is earliest completed by Rangwala in 1987, he introduced a kind of using neural network and acoustic emission technology of tool condition monitoring method; Grabec and Sachse application of artificial neural network of the acoustic emission signal processing, and in the subsequent research, constantly using the neural network technology to acoustic emission signal processing . Barga, such as application error against vividly by the network to simulate the plane frame sample waveform of acoustic emission signal pattern recognition, the results can be of crack extension and friction signal recognition. Yang and Dumont designed a multilayer composite artificial neural network for automatic classification of acoustic emission signals. Eric and others using the acoustic emission signal amplitude parameters in combination with neural network technology, to predict the weld strength of aluminum alloy, has obtained the good effect. James at Johns Hopkins university using neural network to study the complex structure of positioning experiment.

At present, the artificial neural network in the application of acoustic emission technology is aimed at a specific object, lacking a method with universal significance and specificity to acoustic emission signal on the structure of the neural network, so there are still many studies and summarizes to do further.

Correlation analysis of acoustic emission signals

Often in the signal processing the similarity of two signals is to be studied, or a signal after a period of delay after their similarities, in order to realize the signal detection, recognition and extraction and so on. Like the spectrum in the frequency domain analysis, Correlation analysis in time domain apply almost in all areas of signal processing , such as satellite remote sensing ,image processing, radar communication and control engineering, medical and biological engineering, etc. The acoustic emission signal analysis is also often used in correlation analysis technology. For example, you can conduct the similarity analysis of source to the signal of time domain or frequency domain so as to achieve the aim of the acoustic emission source pattern recognition.

Correlation analysis in the acoustic emission signal processing, there is another important application in the continuous analysis of the acoustic emission signal. Commonly used technique measuring time difference between two types of sudden AE wave is not applicable to continuous source of acoustic emission. But crosscorrelation technology applies not only to the discontinuous wave measurement, but also between continuous wave delay measurement, this technique has been successfully applied in pipeline leakage source localization of acoustic emission testing.
(Editor：admin)

Returns

Move>>