·Home ·Table of Contents ·Computer Processing and Simulation

A Phase Deviation Based Split-spectrum Processing Algorithm for Ultrasonic Testing in Coarse Grained Materials Liu Zhenqing, Ta Dean, Liu Xiao (Institute of Acoustics, Tongji University, Shanghai 200092, P. R. China) Contact

Abstract

The performance of an Ultrasonic flaw detection method in coarse grained materials is evaluated by the success of distinguishing the flaw echoes from those echoes scattered by microstructures. The well known Ultrasonic signal processing technique referred as split-spectrum processing (SSP) has proved its success in reduction of such material noise. However, the results of SSP algorithm are not sufficient stable since they are sensitive to the filter band and filter parameters. In this paper, a new SSP algorithm based on phase standard deviation (PSD) is created. The performance has been examined for both computer simulated and experimental data, and compared to other commonly used techniques such as minimization algorithm. The PSD has proved its superior effect and is less sensitive to parameters.

1. Introduction

The coarse grained material due to its intense scattering for ultrasonic, result in that ultrasonic testing signal containing the high level of material structures noise. Therefore, how to improve the signal to noise ratio (SNR) and accurately distinguish the flaw echoes from noise become difficult problems for ultrasonic testing in coarse grained materials. The split spectrum processing (SSP) technique, which was developed in the late 1970s, may be considered as an effective means of improving SNR in ultrasonic flaw detection of coarse grained material [1-3]. However, the results of SSP algorithm are not sufficient stable since they are sensitive to the algorithm's parameter values, so it is very difficult to employ the technique in practical application and dissemination in industry.
In this paper, a new SSP algorithm based on phase standard deviation (PSD) is created, which is less sensitive to processing parameter values, and performs stably. Both the processing results of computer simulated and experimental data proved that the PSD algorithm has superior effect than other relative algorithms.

2. Theory

The theory of SSP and processing course, see Fig.1. First, arm at the spectral range of ultrasonic testing signal, we can assume several band-pass filters, and make the testing signal pass through the bank of band- pass filters, then perform statistical analysis, and signal resuming algorithm. The signal is resulted in new waveform in time region. In the past, several algorithms have been proposed in SSP, the most well known are the methods of minimization (MIN), geometric mean (GM) and polarity thresholding (PT). Previous work in ultrasonic signal processing has shown that the performances of these methods are not sufficient stable such as, when using the same signal resuming algorithm, the number of band- pass filters has considerably impacted the processing results. Therefore, the key of application and dissemination of this technique is to developing a signal resuming algorithm of stable performance.


Fig 1: Theory of SSP. (a): spectrum split; (b): processing course.

The author has theoretically discussed and proved that [4], among the output signals of filters, the phases near the flaw echoes have little varieties. While the coarse grain scattering signals pass through the bank of band-pass filters, the phases have more varieties. So the paper developed a new SSP resuming algorithm based on PSD.
If the received ultrasonic testing signal is r(t), the output signal of SSP band-pass filter bank can be expressed as:


(1)

where N is the number of filters and "T" denotes the transpose, the phase function j (t)can be written
as:

(2)

Therefore, a corresponding to each time instant, the PSD of output signals of different filters can be expressed as:


(3)

where the using of SSP signal resuming algorithm is given by:

(4)

Clearly, this algorithm's ability of enhancing flaw echoes is decides by the property of the phase standard deviation
Std[q _i (t)]. The output signals are enhanced since the PSD is little near the flaw echoes, while it is large in grained scattering, so the signal is suppressed.

Fig 2: The processing course of SSP based on phase standard deviation

The processing course of SSP in the paper developed, see Fig.2. This method operates fast Fourier transform (FFT) for signal, and obtain the spectral range of signal, then assume several narrow band-pass filters (filter-bank). The outputs of filter-bank are sent to PSD to operate signal resume and produce the signal of interesting.

3. Experiment

Fig 3: The computer simulated signal in coarse grained material

In order to examine the performance of the new algorithm, above all, we processed the computer simulated ultrasonic testing signal in coarse grained material, which is shown in Fig.3. The ordinate in Fig.3 is normalized amplitude (the same below). There is a simulated flaw echo at about 10ms~11ms in Fig.3. The results of PSD algorithm are appeared in Fig.4, where Fig.4-1 to Fig.4-4 shows that the results of different filter number N in the spectral range of ultrasonic signal. It is thus clear that the processing results are less dependent on the number of filters. In order to contrast, we used the minimization algorithm to process the signal in Fig.3, the results are shown in Fig.5. The same as the Fig.4, Fig.5-1to Fig.5-2 shows the results of different filter number N. Clearly, this algorithm is very sensitive to the number of filters.






Fig 4-1: Filter number N=8	Fig 4-2: Filter number N=16	Fig 4-3: Filter number N=24	Fig 4-4: Filter number N=32
Fig 4: The results of phase standard deviation algorithm

Fig 5-1: Filter number N=8	Fig 5-2: Filter number N=16	Fig 5-3: Filter number N=24	Fig 5-4: Filter number N=32
Fig 5: The results of minimization algorithm

In parallel, we did experiments to test and verify the new algorithm. The experimental system is shown in Fig.6. Where ultrasonic analyzer is Panametrics 5052UA; The transducer (K&K's K1S-56125 model 45 °oblique probe) worked as the transmitter and receiver. The signals were sent from the ultrasonic analyzer to HP 54600 Oscilloscope and were observed. At the same time, the digitized experimental signals were sent to AST computer and were processed. Fig.7 shows the experimental signal in coarse grained material, the flaw echo can be observed at about 10ms in Fig.7. The processing results of PSD algorithm are given in Fig.8, where Fig.8-1 to Fig.8-4 shows the results of different filter number N from 8 to 32, it also shows that the processing results are less dependent on the number of filters. The processing results of Corresponding to minimization algorithm are given in Fig.9, it also shows that the filter number has a marked effect on the processing result.


Fig 6: Block diagram of the experimental system

Fig 7: Experimental signal in coarse grained material

Fig 8-1: Filter number N=8	Fig 8-2: Filter number N=16	Fig 8-3: Filter number N=24	Fig 8-4: Filter number N=32
Fig 8: The results of phase standard deviation algorithm

Fig 9-1: Filter number N=8	Fig 9-2: Filter number N=16	Fig 9-3: Filter number N=24	Fig 9-4: Filter number N=32
Fig 9: The results of minimization algorithm

4. Conclusion

Cause the results of SSP algorithm are not sufficient stable, a new SSP algorithm based on phase standard deviation (PSD) is created. The performance has been examined for both computer simulated and experimental data, and compared to minimization algorithm. The PSD algorithm has proved its superior effect and is less sensitive to parameters, the SSP's viability of application has been enhanced. The algorithm can be available for use on the industrial scene

Acknowledgment

This project was supported by Natural Science Foundation of China, Grant No. 19574039

References

1. V. L. Newhause, N. M. Bilgutay, ea al. Flaw-to-grain echo enhancement by split-spectrum
Processing. Ultrasonics, 1982; 20(2): 59-68

2. M. G. Gustafsson, T. Stepinski. Split spectrum algorithms rely on instantaments phase
Information - a geometrical approach. IEEE trans. UFFC, 1993; 40(6): 659-665

3. Q. Tian, N. M. Bilgutay. Statistical analysis of split spectrum processing for multiple target
detection. IEEE trans. UFFC, 1998; 45(1): 251-256

4. Z. Q. Liu. Signal enhancement of ultrasonic testing for coarse grained materials and its pattern
   Recognition. Ph.D. Thesis, Tongji University, Shanghai, China; September 1994.

© AIPnD , created by NDT.net

|Home|    |Top|