научная статья по теме A NEW ULTRASONIC SIGNAL PROCESSING SCHEME FOR DETECTING ECHOES OF DIFFERENT SPECTRAL CHARACTERISTICS IN CONCRETE USING EMPIRICAL MODE DECOMPOSITION Общие и комплексные проблемы технических и прикладных наук и отраслей народного хозяйства

Текст научной статьи на тему «A NEW ULTRASONIC SIGNAL PROCESSING SCHEME FOR DETECTING ECHOES OF DIFFERENT SPECTRAL CHARACTERISTICS IN CONCRETE USING EMPIRICAL MODE DECOMPOSITION»

УДК 620.179.16

A NEW ULTRASONIC SIGNAL PROCESSING SCHEME FOR DETECTING ECHOES OF DIFFERENT SPECTRAL CHARACTERISTICS IN CONCRETE USING EMPIRICAL MODE DECOMPOSITION

S. Haddad, A. Bouhadjera, M. Grimes, T. Benkedidah

NDT Lab, Faculty of Sciences and Technology, Jijel University, Algeria

Abstract. In this work, a new signal processing scheme is proposed to improve ultrasonic echoes detection in concrete, a construction material and inhomogeneous propagation medium, in which the backscattering and the grain noise mask the useful signals. Firstly, the time-domain ultrasonic signal is measured. Then the original signal is separated into Intrinsic Mode Functions (IMF) using Empirical Mode Decomposition. Secondly, the Fourier analysis is performed on each IMF. Examining the behaviors of the IMFs in the time and frequency domain, one can judge which IMF represents noise, echoes wave. Then we use nonlinear processing techniques to recombine the selected IMFs. Experimental results are presented and analyzed on a trapezoidal concrete/mortar/cement-paste prism specimen, showing that the proposed technique has an excellent performance. The improvement in defect detection was experimentally verified too on a prism based specimen of mortar with artificial defect.

Keywords: non-destructive testing, Empirical Mode Decomposition, polarity thresholding, Coherent Noise, prism technique.

НОВЫЙ СПОСОБ ОБРАБОТКИ УЛЬТРАЗВУКОВЫХ СИГНАЛОВ

С ИСПОЛЬЗОВАНИЕМ ЭМПИРИЧЕСКОГО РАЗДЕЛЕНИЯ

КОМПОНЕНТ, ПРЕДНАЗНАЧЕННЫЙ ДЛЯ ОБНАРУЖЕНИЯ ЭХОСИГНАЛОВ С РАЗЛИЧНЫМИ СПЕКТРАЛЬНЫМИ ХАРАКТЕРИСТИКАМИ В БЕТОНАХ

С. Хаддад, А. Бухаджера, М. Гриме, Т. Бенкедида Лаборатория неразрушающего контроля, факультет науки и технологии, Джижелъекий универеитет, Алжир

Предложена новая схема обработки у. з. эхосигналов, которая улучшает их обнаружение в бетонах, конструкционных материалах и в других неоднородных средах, в условиях, когда структурный и тепловой шум затеняют полезные сигналы. Сначала во временной области измеряют у. з. сигналы, затем производят разложение исходных сигналов по комплекту значимых функций (intrinsic mode funcion, IMF) с использованием эмпирического принципа и выполняют преобразование Фурье каждой функции IMF. Проанализировано поведение функций IMF во временной и частотной областях, что позволяет решить, какая из них представляет шум, а какая полезный сигнал. Затем авторы используют технологию нелинейного преобразования для разделения выбранных функций IMF. Представлены и проанализированы экспериментальные данные, полученные на образцах отливок их бетона. Показано, что предлагаемый способ обработки дает великолепные результаты. Также на обливке из бетона с искусственной несплошностью экспериментально показано преимущество данного способа обработки.

Ключевые елова: неразрушающий контроль, когерентный шум, методика призмы.

1. INTRODUCTION

In ultrasonic non-destructive testing (NDT) of highly scattering materials, detection of reflected echoes from defects or boundaries is difficult due to the masking effect of the structural noise. The most frequently used signal-processing method for noise reduction is time averaging but it can reduce only the incoherent content of the signal noise. This is not enough for testing concrete because the remaining noise, which is the coherent (or scattering) noise produced by

Sofiane HADDAD, BP 157 Taher 18002 W. Jijel, Algeria. Tel: +213(0)550895598. E-mail: sof_had@yahoo.fr

backscatter of the propagating ultrasound, is still irresistible. In this work, empirical mode decomposition (EMD proposed by Huang et al. [1]) is used to deal with detecting complex multiple targets in ultrasonic applications. EMD is one of the most advanced concepts considered in digital signal processing and focused on processing nonlinear and nonstationary processes. In comparison with the classical Fourier analysis and wavelet algorithms, EMD has a very high extent of adaptation to processing various nonstationary signals. Furthermore, it does not impose any serious restrictions on the harmonic nature of basis functions. The key role is played by empirical mode decomposition, which allows any complicated signal to be decomposed into finite and a usually very small number of empirical modes (IMFs—Intrinsic Mode Functions), each containing information about the initial signal.

2. EMPIRICAL MODE DECOMPOSITION

The Empirical Mode Decomposition (EMD) is a signal processing technique capable of extracting all the oscillatory modes present in a signal at different length scales. EMD decomposes a time series into components by empirically identifying the physical time scales intrinsic to the data. Each extracted mode, named Intrinsic Mode Function (IMF), is symmetric with respect to zero, has a unique local frequency, and different IMFs do not share the same frequency at the same time.

Fig. 1. Scheme of the proposed method in order to process the ultrasonic signals.

The decomposition method [3] involves the following steps (Fig. 1):

Step 1: Given a signal x (n), finding the local maxima of the signal, all the maxima are connected by a cubic spline line as the upper envelope. Then finding the local minima of the signal, all the minima are connected by a cubic spline line as the lower envelope. The mean of the upper envelope and the lower envelope is designated as m1 (n),

h1(n) = x(n) - m1(n). (1)

If h1(n) satisfies all the requirements of the IMF, it is the first IMF.

Step 2: if h1(n) does not satisfy the conditions of the IMF, it is treated as the datum, and Step 1 is repeated, m11(n) is the mean of the upper and the lower envelopes of h1(n), then

h11(n) = h1(n) - m11(n).

(2)

If h11(n) satisfies all the requirements of the IMF, it is the first IMF. If h11(n) does not satisfy all the requirements of the IMF, Step 1 is repeated continually up to k times until h1k(n) is an IMF, that is,

hu(n) = hUk_1)(n) - mUk_1)(n). (3)

It is designated as the first IMF component c1(n) from x(n), that is, c1(n) = = h1k(n). Then c1(n) is removed from x(n) to obtain the residue r1(n), that is,

r^n) = x(n) _ C1(n). (4)

The residue r1(n) is treated as the new datum. Then Steps 1 and 2 are repeated to obtain the second IMF c2(n). This procedure is repeated to obtain all the IMFs, and the results are

r(n) = r_1 (n) _ c(n), i = 2, 3, ..., l. (5)

The decomposition stops as any of the following predetermined criteria is achieved: (1) Either the component cl(n) or the residue rl(n) becomes so small that the residue is less than the predetermined value of the substantial consequence; (2) The residue rl(n) becomes a monotonic function from which no more IMF can be extracted. Thus, the original data are the residue plus the sum of the IMF components:

' i

x(n) = rl(n) + X c>'(n)- (6)

i=1

To guarantee that the IMF components retain enough physical sense of amplitudes and frequency modulations, a criterion for stopping the EMD process is determined by SD,

SD = X

Kk-i)(n)- hik (n)

n=0

K

1(k-1)(n) '

A typical value for SD can be set between 0.2 and 0.3 [1].

(7)

3. BASIS OF THE PROPOSED SCHEME

Empirical Mode Decomposition is an emerging new technique of signal decomposition having many interesting properties. In particular, EMD can be applied to nonlinear; nonstationary noisy signals and does not require any prior knowledge on the nature and number of modes embedded in a signal. The typical EMD application generates a number of intrinsic modes from one single signal: this makes EMD a powerful technique that can be used when dealing with multiple targets echoes. In order to improve the detection of multiple reflected echoes with different spectral and temporal characteristics using empirical mode decomposition, five steps need to be followed:

Step 1: Decompose the ultrasonic signal by EMD to obtain the finite IMF components;

Step 2: Apply FFT to each IMF component to obtain the frequency information;

Step 3: Calculate the energy of each IMF;

Step 4: Select IMF components that have the maximal energy;

Step 5: Recombine the selected IMF components using a non-linear processor.

In this paper we have used the polarity thresholding (PT) to recombine the selected IMFs. The PT output can be expressed as

Jy(t) = max(|IMF,.(t)|) (IMF,.(t)> 0 or IMF,.(t)< 0, j = 1,2,...N) [ y(t) = 0 (otherwise)

4. EXPERIMENTAL RESULTS 4.1. Experiment 1

This experiment was performed using the prism technique [2] to detect the multiple echoes reflected from a sample of a trapezoidal concrete (mortar) cement-paste prism whose dimensions are 21x10x7.4(lc = 2+lm = 2.2+lcp = 3.2) cm (Fig. 2) with 0.5 W/C ratio, 0.5 C/S ratio and 9 % aggregate with 0.5 maxi-

Fig. 2. General view of the measuring system: experiment 1.

mum particle size for the concrete layer, 10 days after making. Measurements were made in pulse-echo mode in the longitudinal direction using immersion transducer with 0.5 MHz center frequency and 0.75 inch diameter. The main objective was the detection of reflected echoes from different interfaces (E1: water-concrete, E2: concrete-mortar, E3: mortar-cement paste and E4: cement water). Consequently, the received signal is processed first using EMD to obtain a set of IMFs; IMFs are recombined nonlinearly to improve echoes visibility.

4.2. Experiment 2

In this experiment, we examine a trapezoidal-prism of mortar with 0.5 w/c ratio and 1 s/c ratio. The mortar is about 13 days old. In the mixing phase we have added a small piece of aluminium (Fig. 3) and we considered it as a defect in the

Fig. 3. General view of the measuring system: experiment 2.

specimen (changing in the acoustic impedance). Measurements were made in the pulse-echo mode in the longitudinal direction using immersion transducer with 2.25 MHz center frequency and 0.5 inch diameter.

4.3. Discussions and remarks

The resulting output shown in Fig. 4 clearly extracts the last three echoes (E2, E3 and E4) from the background grain noise in the first IMF. The third IMF con-

Original signal

Для дальнейшего прочтения статьи необходимо приобрести полный текст. Статьи высылаются в формате PDF на указанную при оплате почту. Время доставки составляет менее 10 минут. Стоимость одной статьи — 150 рублей.

Показать целиком