Damage detection of pipeline multiple cracks using piezoceramic transducers

2.1. Specimen design

One section of a steel pipeline sample was used as the specimen in this experiment. The pipeline was constructed from Q235 steel. The pipeline has a length of 100 mm with an outer diameter of 101 mm and an inner diameter of 80 mm. Two artificial cracks, with depth increasing from 0 mm to 9.0 mm at an interval of 1.5 mm, were created by mechanical cutting method. The second crack was created after the first one reached its full depth. Seven different operating conditions were tested for each crack. Table 1 shows each operating condition with its different crack depth. It should be noted that the crack depth was increased by 1.5 mm for each operating condition starting from Condition 1, and the crack width was kept at 1.2 mm throughout. PZT-5H was used to as transducers to identify the pipeline multi-cracks damage. The main properties of PZT patches are presented in Table 2. Fig. 1 shows the locations of PZT1, PZT2, PZT3 and the artificial cracks. Fig. 2 shows the locations of PZT3 and PZT4. It should be noted that the artificial crack I was positioned halfway between PZT1 and PZT2 and the artificial crack II was positioned halfway between PZT2 and PZT3.

Fig. 1Locations of PZT sensors and cracks

Fig. 2Locations of PZT3 and PZT4

2.2. Experimental setup and testing procedures

Fig. 3 shows the experimental setup, which mainly includes the pipe specimen with PZT-5H patches, an impedance analyzer (Agilent 4294A), a function generator (Agilent 33120A), a data acquisition system (NI USB-6363), and a host PC. When the test is carried out using the electromechanical impedance method, through analyzing the preliminary results measured by the impedance analyzer, the effective frequency range was chosen to be 60-90 kHz. During the tests, received electromechanical impedance signals of PZT1, PZT2, PZT3 and PZT4 in each operating condition were respectively collected and analyzed. When the test is carried out using the stress wave method, a sweep sine wave signal generated by the function generator was used, and the sensitive frequency interval (60 kHz-200 kHz) was determined in this research. This sweep sine signal was then used as the excitation source. During the tests to detect crack I, PZT1 was used as an actuator and PZT2, PZT3, PZT4 were used as sensors. During the tests involving crack II, PZT2 was used as an actuator and PZT1, PZT3, PZT4 were used as sensors. The process was automated using the Labview software for signal generation and data acquisition. For each damage condition, the electromechanical impedance method is used first, followed by the stress wave method.

Fig. 3Experimental setup

3.1. Experimental results based on the electromechanical impedance method

During the test of crack I, seven operating conditions with different crack depths (0 mm-9.0 mm) were investigated. Table 1 depicts each operating condition with its corresponding crack depth. All PZT transducers are used. Test of crack II was performed after the test of crack I, and the operating conditions and monitoring method remained the same. The test results show that the electromechanical impedance changes are the most significant from 60 kHz to 90 kHz. Fig. 4 reflects the impedance spectrum curves of PZT3 in the tests of crack I and crack II. The entire monitoring results received from PZT1, PZT2 and PZT4 are consistent with those received from PZT3 and are not plotted in this paper. As shown in Fig. 4, with the increase of the crack depth, both impedance peaks and peak frequencies decrease. This explains that the electromechanical impedance method is sensitive to damages.

Fig. 4The impedance spectrum curve of a), b) PZT3 and c), d) PZT4

a) Crack I

b) Crack II

c) Crack I

d) Crack II

3.2. Experimental results based on the stress wave method

In the active sensing approach, one PZT transducer is used as an actuator to generate the desired wave to propagate through the host structure, and other distributed PZT transducers are used as sensors to detect the wave response. Cracks or damages inside the structure act as a stress relief in the wave propagation path. The amplitude of wave and the transmission energy will decrease due to the existence of cracks or damages. On the pipeline segment, one PZT was used as the actuator in order to generate the swept sine wave signal. Meanwhile, three PZTs were set up at different locations on the pipeline as sensors to receive the excitation signal from the actuator.

Operating conditions of identification test based on stress wave method are shown in Table 1. During the test of crack I, PZT1 was used as an actuator and PZT2, PZT3, PZT4 were used as sensors. While during the test of crack II, the actuator was PZT2, and the sensors were PZT1, PZT3 and PZT4. The period of the received signal was 2 seconds. Since the amplitude of the peaks of stress wave signals are concentrated in 1.6-2.0 seconds, the stress wave shown in this paper are in this interval (1.6-2.0 seconds). The received stress wave signals of PZT2 and PZT4 during the test of crack I are shown in Fig. 5(a) and Fig.6(a), respectively. In order to compare with identification results of crack I, the received stress wave signals of PZT3 and PZT4 during the test of crack II are shown in Fig. 5(b) and Fig. 6(b). And the received stress wave signals of PZT1 during the test of crack II is shown in Fig. 7. Fig. 5, Fig. 6 and Fig. 7 show that the peak voltages received by PZT sensors decreases with the increase of crack depth. During the test of crack I, the received peak amplitude from PZT2 is greater than those from PZT3 and PZT4, while the received peak amplitude from PZT3 is greater than that from PZT4 during the test of crack II, which concludes that the closer to the excitation source, the more sensitive of received signals. During the test of crack II, the received amplitude of signals peak of PZT1 does not change significantly. It is found that the received signals will not change unless a new damage appears between the excitation sources (PZT2) and the sensor (PZT1). As a result, when a sensor array is used, the location and severity of cracks can be determined by monitoring the sensor signal changes while alternating the actuating PZT actuator and PZT sensors.

Fig. 5The stress wave monitoring results of PZT2 and PZT3

a) PZT2 (Crack I, OC1-7)

b) PZT3 (Crack II, OC1-7)

4.1. Damage index based on the electromechanical impedance method

For this study, the Root-Mean-Square Deviation (RMSD) of the impedance of the PZT patches used as the damage index or indicator, which was first proposed by Giurgiutin and Rogers [26]:

where $y_i$ is a post-damage impedance signature at the $i$th measurement point, and $x_i$ is a health value at the $i$th measurement point.

Fig. 6The stress wave monitoring results of PZT4

a) PZT4 (Crack I, OC1-7)

b) PZT4 (Crack II, OC1-7)

The experimental results of the impedance based RMSD indices are shown as Fig. 8. As we can see that, the RMSD indices for all PZT sensors increase with the crack depth, and the test results of crack I and crack II exhibit the same trend. The test results of crack I indicate that the PZT1 and PZT2 have higher damage index values than those of PZT3 and PZT4, which indicates that PZT1 and PZT2 are closer to the crack I than PZT3 and PZT4. The test results of crack II indicated that the PZT2 and PZT3 have larger damage index values than those of PZT1 and PZT4, which indicates that PZT2 and PZT3 are closer to the crack II than PZT1 and PZT4. In summary, the experimental results show that the largest damage index value is always associated with sensor that has the shortest distance to the crack, which can be used to quantitatively analyze the pipeline crack damage.

4.2. Wavelet packet energy damage index based on the stress wave method

The crack in the stress wave propagation path functions as a stress relief. Furthermore, the loss of energy received by the PZT sensors is correlated with the severity of the crack. These phenomena are then quantified with wavelet packet analysis, which is used as a signal-processing tool for analysis [27]. The basic principles of this analysis technique are given as follows.

Fig. 7The stress wave monitoring results of PZT1

a) PZT1 (Crack II, OC1-4)

b) PZT1 (Crack II, OC5-7)

Fig. 8Damage index RMSD based on electromechanical impedance monitoring results

a) Damage index RMSD of crack I

b) Damage index RMSD of crack II

In the proposed health monitoring algorithm, the sensor signal $V$ is decomposed by an $n$-level wavelet packet decomposition into $2^n$ signal subsets $\{X_1,\mathrm{}{X}_2,\cdots ,X_{2^n}\}$ and $j$ is the frequency band. The decomposed subset $X_j$ is written as:

where $m$ is the number of sampling data, and the energy of the decomposed signal at time index $i$ can be defined as:

The energy vector at time index $i$ can be defined as:

Based on the calculation of energy vectors ($E_i$) (Eqs. (3)-(5)), the crack severity index for the sensor signal at time index $i$ can be expressed as:

where $E_{1,j}$ represents the energy level when this no damage to the specimen.

When the crack width is kept as 1.2 mm, the wavelet packet energy loss $I\left(i\right)$ (abbreviation WPEL) can be an approximated indicator of the crack severity. Based on the stress wave monitoring data, Fig. 9 shows that WPEL indices of crack I and crack II in each operating condition. It can be seen that the WPEL indices increase with the crack depth. Therefore, the pipeline crack damage can be described by the WPEL index. In the tests of two cracks, the PZT sensor that is nearest to the crack has the highest damage index value, which indicates that the identification results were directly correlated to the locations of the PZT sensors with respect to the crack. In the process of increasing the depth of crack II, the WPEL index of PZT1 did not change, and this shows that crack II damage has no effect on stress wave received by PZT1, from which we can conclude that no new damages are formed between PZT1 and PZT2.

Fig. 9Damage index based on stress wave

a) Damage index of crack I

b) Damage index of crack II