High Precision Ultrasonic Monitoring

 

High Precision Ultrasonic Monitoring Alan Puchot, Adam Cobb, PhD, Glenn Light, PhD Southwest Research Institute San Antonio, TX 78228 Ali Minachi, PhD Aramco Dhahran, Saudi Arabia Abstract SwRI with Aramco has developed a new concept for monitoring corrosion in  pipelines  -  The  The inspection transducer is bonded to the pipe in transducer hold down fixtures that is magnetically held in place that will ultimately be connected to a multi-channel, wireless wireless data acquisition and analysis system. To accomplish this goal, a number of parameters had to be studied. This paper discusses the work conducted to accomplish the initial goal.

1. INTRODUCTIO INTRODUCTION N Over the lifetime of a gas pipeline, the pipe wall can thin or otherwise degrade as a result of corrosion and fatiguing caused by exposure to the surrounding environment. In the humid environments such as the Saudi Arabian Persian Gulf, pipelines are particularly susceptible to corrosion due to water collection on the pipe interior and salt buildup on the pipe exterior. Many refineries and pipeline companies practice randomized spot inspections of their pipelines to ensure the wall condition meets their safety regulations. Because these inspections are usually conducted by a field technician, it is difficult to acquire consistent thickness measurements over long periods of time. This often leads to inaccurate trending. To improve the consistency of these measurements, Aramco and SwRI worked together to develop initial specifications and concept designs for ultrasonic monitoring systems on their pipelines to measure with high precision the remaining wall thickness at critical locations. Hardware requirements for for a high precision precision ultrasonic monitoring monitoring system were investigated and any limiting factors that would that would affect accuracy were identified. In addition a conceptual design for a high precision, remote monitoring system has been developed and is in the process of development. The initial specifications specifications for a high precision ultrasonic monitoring monitoring system were identified as the following: System must resolve carbon steel pipe thickness measurements in the range of 2 mm to 30 mm to an accuracy of ±0.5 mm. Probes must operate continuously over the full environment temperature range, including temperatures up to 120ºC and specified measurement accuracy must be maintained over the full temperature range. System must accommodate cable lengths up to 50 meters between the pulser and the probes without affecting measurement accuracy.

 

  Pulser instrument must be capable of driving multiple probes (20 probes was w as set as the goal). System will include a pulser that will drive multiple probes, a data acquisition system, and an autonomous data analysis station. Base station will consist of a data management system s ystem for controlling multiple local systems, and user driven analysis software. Base station will control the local stations through a wireless communication technology. A series of tests were conducted to determine how best to meet these spec specifications. ifications.

2. EXPERIMANTAL WORK CONDUCTED The experimental work conducted investigated (1) transducer type and frequency, (2) cable length, and (3) temperature temperature effects. To accomplish the ultimate goal, a method to  permanently and robustly attach the transducer to the pipe surface and a computer algorithm function allowing the computer to determine the remaining wall were developed.

2.1 Transducer Type and Frequency Frequency During the initial evaluations, SwRI considered three different types of normal incidence ultrasonic contact probes for the monitoring system   system   single single element probes, dual element  probes, and delay line probes. Initial testing was conducted on a flat carbon steel step wedge with thicknesses of 2.54 mm to 12.7 mm in 2.54 mm increments to evaluate the effectiveness of each transducer type at measuring measuring wall thicknesses. During these tests, measurements were taken with two dual element transducers with frequencies of 2.25 MHz and 5 MHz, and with several single element longitudinal transducers with a frequency range of 2.25 MHz to 15 MHz and diameters of 6.4 mm, 9.5 mm, and 12.7 mm. Tests were also conducted with delay lines added to the normal incidence transducers.

The transducers were driven in pulse-echo mode with a Panametrics 5072 pulser and the waveforms were collected with a Tektronix TDS 2024B oscilloscope. The cable length  between the pulser and the probe was negligible (1.2 meter). A function was written in MATLAB to analyze the waveforms collected and to report the thickness of the measured test sample. The MATLAB function identified the back wall signal and each subsequent multiple reflection in the waveform and then measured the time difference between each adjacent back wall signal. The time difference measurement was conducted either through cross correlation of the adjacent back wall signals, or by a peak detection algorithm. The measured time differences were used to calculate the thickness of the sample based on the material velocity. Additionally, the function returned information regarding the center frequency and bandwidth of the back wall signal, as well as the signal attenuation between each back wall. An example of the

 

resulting annotated waveform is shown in Figure 1. Table 1 provides the resulting data from each measurement taken during the initial thickness test.

Figure 1. Waveform output of MATLAB thickness measurement function. The first identified back wall is highlighted in red and subsequent reflections are highlighted in green. The function reported a material thickness of 2.55 mm (real thickness of 2.54 mm). Table 1. Transducer performances from the initial thickness measurement test Wedge Thickness (mm)

12.7 10.16 7.62 5.08 2.54 12.7 10.16 7.62 5.08 2.54 12.7 10.16 7.62 5.08 2.54

Reported Percent Error Center Thickness Frequency (mm) (MHz) 2.25 MHz Transducer, 9.5 mm diameter d iameter 12.56 1.10% 2.5 10.02 1.38% 2.5 7.47 1.97% 2.4 4.92 3.15% 2.5 Signal overlap too significant to resolve back wall. 5 MHz Transducer, 12.7 mm diameter 12.67 0.24% 4.4 10.14 0.20% 4.5 7.61 0.13% 4.7 5.1 0.39% 4.2 2.55 0.39% 3.9 10 MHz Transducer, 6.4 mm diameter 12.73 0.24% 7.7 10.19 0.30% 9.4 7.59 0.39% 7.8 5.06 0.39% 7.9 2.53 0.39% 9

Average Attenuation (dB)

3.2 2.5 1.9 1.4

2 1.8 1.7 1.3 1.7 2.8 2.9 2.9 2.1 1.3

The 2.25 MHz transducer performed very poorly in this test. The waveforms collected with this probe tended to be very noisy, and typically only 2 reflected back wall signals could be clearly detected. At the smallest wedge thickness, the 2.25 MHz probe was not

 

able to resolve the back wall from the initial pulse, and the second reflection had become convoluted with the first reflection. Figure 2 shows a typical thickness measurement recorded with the 2.25 MHz probe.

Figure 2. Thickness measurement recorded with the 2.25 MHz transducer on the 7.62 mm thick sample. Only two back wall reflections are clearly present, and an unusually un usually high level of noise is detected. Both the 5 MHz and 10 MHz transducers returned clear back wall reflections and multiple reflections from a thickness of 12.7 mm to 2.5 mm. Figure 3 shows an example measurement conducted on a 7.62 mm thick steel block with the 10 MHz probe. Based on this test, it appeared that a single element longitudinal transducer between the frequency of 5 MHz and 10 MHz would be capable of making pipe thickness measurements within the provided thickness range and accuracy specification.

 

Figure 3. Thickness measurement recorded with the 10 MHz transducer on the 7.62 mm thick sample. Multiple back wall signals are detected, and the noise level is low. 2.2. Evaluation of Transducer Cable Length On Measurement Accuracy

For this cable test, the thickness measurement was and conducted on the 12.7mm A thick step using various lengths between between the transducer the pulser instrument. baseline cable length of 1.2m was used and then additional lengths of 2m, 5m, 10m, 20m, 30m, and 50m of RG-174 cable were connected to the existing 1.2 m cable. Otherwise, all of the instrumentation and instrument settings remained the same. Waveforms were recorded for each combination of probe frequency (2.25 MHz, 5 MHz, and 10 MHz) and cable length. In order to conduct a reasonable evaluation of the effects of the coaxial cables on the transmitted waveforms, the anticipated effect of the coaxial cable on the signal was first modeled using a typical electrical model of a coaxial cable. Each unit length of the cable is noted to contribute some series inductance and some parallel capacitance to the terminating load. This model can be interpreted interpreted as a series of low pass filters, where each differential length contributes to the order of the low pass filter. From this interpretation, it is apparent that higher frequencies will be more heavily attenuated by the coaxial cable, and longer cables will obviously have greater effect on the signal it carries. Very long cables may also cause broadening of higher frequency signals. Additionally, the coaxial cable will act as a time-delay line, so longer cables will tend to phase shift the reflected signal relative to the transmitted pulse. However, because the only critical measurement for the thickness measurement system is time-of-flight, none of these effects may matter as long as (1) the reflected signal amplitude remains clearly detectable above the noise level, (2), the probe center frequency does not become heavily attenuated, and (3) the time difference between two  back wall signals, or a detectable front wall and back wall signal, does not shift significantly enough to prevent accurate thickness measurements of thicknesses greater than 2mm with frequencies of 5 and 10MHz. The data acquired from these tests are summarized in Table 2. Table 2. Transducer performances from cable length test Cable Reflected Signal First Average Reported Length Center Reflection Attenuation Thickness (m) Frequency (MHz) Amp. (V) (dB) (mm) 2.25 MHz Transducer, 9.5 mm diameter 0 2.47 7.29 3.2 12.56 2 2.47 7.30 3.4 12.55 5 2.64 7.30 2.9 12.67 10 2.30 7.22 4.0 11.47 20 2.30 7.31 4.1 12.64 30 2.30 7.22 4.9 12.58 50 2.47 6.91 6.1 12.60

Measurement Error (mm)

0.14 0.15 0.03 1.23 0.06 0.12 0.10

 

0 2 5 10

4.36 4.26 4.12 3.52

20 30 50

4.23 3.85 4.12

0 2 5 10 20 30 50

7.90 8.01 7.08 7.14 6.79 6.76 6.35

5 MHz Transducer, 12.5 mm diameter 7.36 2.0 7.31 1.9 7.29 2.0 5.49 2.1

6.00 2.2 4.57 2.0 2.67 2.0 10 MHz Transducer, 6.4 mm diameter 6.37 2.8 6.34 3.0 5.78 3.1 5.33 3.2 4.57 3.1 3.75 3.3 2.47 3.2

12.67 12.68 12.67 12.68

0.03 0.02 0.03 0.02

12.65 12.69 12.70

0.05 0.01 0.00

12.74 12.74 12.74 12.74 12.74 12.74 12.74

0.04 0.04 0.04 0.04 0.04 0.04 0.04

Table 2 indicates that both the 5 MHz probe and the 10 MHz probe were able to return accurate and consistent thickness measurements regardless of the length of the cable connecting the probe to the pulser instrument. The same consistency is not observed on the 2.25 MHz probe. This may have had less to do with the coaxial cable itself and more to do with the high level of acoustic noise and higher signal attenuation observed on this  probe. Because the back wall multiple reflections of the 2.25 MHz probe were not significantly larger in amplitude than the acoustic noise, the noise distorted the multiples, which reduced the measurement accuracy. Although the 5 MHz and 10 MHz probes tended to produce thickness measurements with much greater accuracy, they were also more significantly affected by increasing cable length. The signal attenuation in the sample for both transducers was typically around 2 to 3 dB. In contrast, the 5 MHz signal experienced 0.09 dB/m of attenuation in the cable, and the 10 MHz probe experienced 0.11 dB/m attenuation. This resulted in significantly significantly smaller signal amplitudes at the maximum cable length. This attenuation will not be a  problem because even at the maximum cable length, the reflections acquired on the 5 MHz and 10 MHz probes were much larger in amplitude than the noise level. The test also demonstrated the tendency of the coaxial cable to act as a lowpass filter. Regardless of the length of the cable, the center frequency of the 2.25 MHz probe appeared to remain unaffected. With the exception of the waveform acquired with the 10 m cable, this was generally true of the 5 MHz probe as well, but the center frequency of the 10 MHz probe was noted to drop by nearly 1.6 MHz from the acquisition with the shortest cable length to the acquisition with the longest cable length. This result indicates that the thickness measurement system will either need to use a probe with a center frequency less than 10 MHz, or a coaxial cable with lower capacitance must be selected. The increased cable length causes a time shift between the spike pulse of the pulser and the initial pulse of the transducer. transducer. However, the relative relative time measurement between between the

 

front surface and the first and subsequent back wall multiple signals does not change. as illustrated in Figure 4.

Phase Shifts of 10MHz Signal from 12.7mm Thick Block with Increasing Cable Length

0.2µ  

0.2µ   Figure 4. Comparison of time-of-flight measurements of the 10 MHz waveform for increasing cable lengths. The upper right plot shows the acquired waveform with no additional cable. The bottom left plot shows the first back wall reflection for all acquisitions, and the bottom right shows the second back wall reflection for all acquisitions. Both signals are noted to shift out in time by roughly 10 nanoseconds for each additional meter of cable. From Figure 4, a time delay of approximately 10 nanoseconds for each additional meter of cable was me me              and all subsequent signals. This delay correlates directly to the time it takes for the electrical signal to travel up and down the length of the cable and is independent of the transducer frequency. This time delay will have to be accounted for in the thickness measurement system if the system does not use a probe capable of capturing multiple  back wall signals. As long as the selected probe can capture either multiple back wall reflections or a back wall and a front wall reflection, adding a long cable between the transducer and the pulser will not produce a measurement error in excess of the error specified.

 

2.3. Evaluation of sound velocity change due to temperature on thickness measurement accuracy.

On a typical pipeline, the probe is expected to experience variations in temperature  between 0ºC and 120ºC due to climate conditions. The probe must be able to both operate for long periods of time at temperatures up to 120ºC and return thickness readings with an accuracy of to ±0.5 mm for pipe walls up to 30 mm thick regardless of the pipe temperature. If the change in velocity as a function of temperature is small enough in carbon steels, it may be possible to ignore the variation of apparent thickness due to temperature change. Otherwise, the measurement system will need to account for temperature variation in the thickness calculation. To determine the significance of the velocity variation over this temperature range, a test was conducted to measure the longitudinal wave velocity v elocity change in 1018 carbon steel. A 5 MHz, 9.5 mm diameter transducer with medium damping characteristics was bonded to TM TM a carbon steel cylinder using a film of 3M  Scotch-Weld  DP125 epoxy. This epoxy epox y was used because it has demonstrated good acoustic transmission properties over this temperature range. The cylinder measured 32 mm in diameter and 40.50 mm m m in thickness. The top and sides of the cylinder was wrapped in insulation and the cylinder was placed on a hot plate. A waveform was acquired at room temperature (22.2 ºC). The waveform included the first back wall reflection and five back wall multiples. The temperature was allowed to gradually increase, and waveforms were acquired periodically. Prior to each acquisition, the temperatures of the cylinder top surface and bottom surface were recorded and averaged. The last waveform was recorded once the t he cylinder reached an average temperature of 100.6ºC. The transducer was observed to fail after exceeding a temperature of 110ºC. The time of flight at each temperature was determined using the thickness measurement algorithm, and the wave velocity was calculated for each temperature. The plot of wave speed as a function of time is shown in Figure 5.

 

Figure 5. Velocity as a function of temperature in 1018 steel. At 0ºC, 0 ºC, the wave speed is -4                

 

-4

The rate of velocity change was calculated to be 8.015*10       steel, and the velocity at room temperature (22ºC) was measured to be approximately                  apparent thickness change of 0.41 mm will occur between a waveform acquired at room temperature and a waveform acquired at the maximum specified temperature. In addition to the other measurement errors inherent to the system, s ystem, this shift may exceed the tolerable error specified, but only in the case of the greatest pipe thickness. If the specified accuracy is critical even at the most extreme case of the greatest pipe thickness and the largest specified temperature change, then it may be necessary to include temperature measurement capability on the thickness measurement system. 2.4. Field test of commercially available transducer under environmental conditions

A fixture (called the transducer hold down fixture or THDF) was designed to hold the  probe in place on pipes with a range of diameters. The THDF consisted of two magnetic shoes that anchored the probe against the pipe and a crosspiece that attached to the probe casing. The THDFs for the D790-SM transducer and a single transducer with a delay line are shown in Figure 6. The probe face is coupled to tthe he pipe surface with a film of epoxy. The epoxy served only as an ultrasonic couplant and was not intended to mechanically support the probe.

 

Figure 6. Transducer Hold Down Fixture for Dual element transducer (left) and Single Element Transducer with Delay Line (right) held against the test pipe Several waveforms were acquired from the transducer over a course of one month with a sampling rate of roughly waveform every three days. All of the waveforms were acquired using the same instrument settings. It is worth noting that over the course of the test, the  probe was subjected to rain, as well as a large temperature change driven by a cold front. front. The maximum time variation between the back wall signals of any an y two waveforms recorded over the course of the test did not exceed 30 ns, which equates to less than 0.09 mm in wall thickness in carbon steel. The primary cause of the variation was minute changes in wave speed. This cause was confirmed by noting that the changes in absolute  position of the signals in each waveform scaled proportionally to the change change in absolute  position of the pipe back wall signal. In essence, the signals remained basically unchanged over the course of the test. More testing is needed to better characterize the robustness of this arrangement over time. 2.2. Evaluation of Thickness measurement accuracy to simulated defect growth

Although testing had been conducted to demonstrate thickness measurements over the full thickness range on smooth, even surfaces, these tests did not prove that the developed monitoring approach and associated analysis function could detect actual defect growth. A test was devised to simulate a more real-world case. A 10 MHz, 9.5 mm diameter transducer was attached to a 15 mm long Rexolite delay line with epoxy and then installed in the THDF that was attached to a 9.5 mm thick carbon steel plate as shown in Figure 7, and a film of epoxy epox y was added between the delay line and the plate.

 

 

  Figure 7. Transducer with delay line attached to carbon steel plate with the designed  probe fixture. The plate measured 9.5 mm thick. The round bottom hole under the plate is shown on the right hand side of the Figure. The hole was cut with a 32 mm diameter  ball end-mill. At this point, the hole measures measures roughly 5.1 mm deep. Gouges from the dremel tool can be seen throughout the hole surface. The plate was then mounted in a milling machine with the transducer fixture facing down. A 32 mm diameter ball end-mill was centered directly over the transducer. The  pulser andand oscilloscope settings were adjusted so that the first front surface surf signal did not saturated so that multiple front surface reflections were visible. Thisace waveform was recorded, and then a second waveform was recorded with a significantly reduced voltage range. In later acquisitions, recording the waveform with a reduced voltage range would allow low level signals to become more apparent. The ball end-mill was driven 0.508 mm into the plate, and two more waveforms were recorded. A dremel tool was then used to gouge the surface of the resulting pit, and two more waveforms were recorded. This procedure continued with the depth of the hole increasing in 0.508 mm increments until the hole reached a total depth of 2.54 mm. The test then continued in depth increments of 1.27 mm until the hole reached a total depth of 5.08 mm. Towards the later part of the test, as the surface area of the hole increased, the gouges added with the dremel tool became more pronounced. p ronounced. The final hole after the dremel tool was applied may be seen in Figure 7. This test, as well as other preliminary tests conducted during the later stages of the  project, demonstrated that under most cases, looking for multiple back wall reflections in a waveform would be a highly unreliable approach to making wall thickness measurements. Based on the results of this test, the previous thickness measurement algorithm was abandoned, and a new approach was developed. The new algorithm relies on the delay line attached to the transducer to detect both one front wall signal and one  back wall signal. It generates an envelope of each of those signals and then defines the center of the pulse width based on these envelopes. The algorithm then uses the separation of the two pulse widths to estimate material thickness. Because the algorithm uses the pulse widths of the signals and does not need information regarding the signal  peaks, the waveforms with the higher sensitivity s ensitivity settings were used. Figure 8 illustrates the performance of this approach.

 

Figure 8. Comparison of the estimated remaining wall thickness under the probe against the actual thickness of the remaining wall. The points shown represent estimates derived from the waveforms taken with the frontwall reflection saturated. Once defect growth  began, all estimates remained within 0.5 mm of the actual thickness. thickness.

 

Overall, the final algorithm performed very well at making thickness measurements. Further revisions will be needed to harden the code against failure, and further testing will be needed to verify the performance of the algorithm. In addition, it was determined that the Rexolite delay line attached to the transducer significantly attenuated the center frequency of the transmitted signal. Therefore, a better delay line material is needed.

4. Conclusions  The results of the testing conducted in this project demonstrate that it is feasible to design an autonomous ultrasonic monitoring system capable of measuring pipe wall thicknesses from 2 mm to 30 mm. Based on the work conducted, the following conclusions were reached. 1.  A single element normal incidence transducer with an attached delay line is recommended. 2.  The probe frequency should be between 5 MHz and 10 MHz with an ideal center frequency of 7 MHz, and the probe diameter will be 12.7 mm nominally. 3.  Rexolite has been used in the testing for a delay line, but this material is not optimum, and to achieve the accuracy needed at the lowest thickness range, a different delay material is needed. More work is needed to optimize the delay line material, but it is anticipated that a metallic material such as aluminum should  be used. 4.  Cables up to 50 m in length may be connected between the transducer and the  pulser instrument without the need of a preamplifier preamplifier at the transducer.

 

TM

TM

5.  Transducers can be attached to inspection surfaces using 3M   Scotch-Weld   DP125 epoxy in conjunction with a mechanical fixture to effectively hold the  probe in position on the pipe. The mechanical fixture will maintain the contact  pressure between the probe and the pipe, thus ensuring measurement consistency, and will protect the epoxy bond between the probe and the inspection surface from damage. 6.  A robust algorithm was developed for calculating the remaining wall when a single transducer with a delay line is used.

Although the fundamental requirements of high precision ultrasonic system are now understood, further study is still needed to finalize the hardware and software specifications of the autonomous monitoring system.