1.  Introduction

Heat recovery steam generators (HRSGs) are susceptible to various degradation mechanisms—Flow Accelerated Corrosion (FAC), pitting and fatigue cracking. In order to harvest maximum heat from the exhaust gas, thousands of tubes are stacked together with narrow gaps between tubes.  The compact design imposes access restrictions for inspection and repair. Failures are often unpredicted, resulting in undesirable, forced outages.

In this paper, we describe a technique to internally inspect HRSG finned tubes via the header’s inspection ports. A Near Field Array (NFA) is used to generate C-scans, Absolute traces and Lissajous patterns characterise the internal integrity of HRSG finned tubes.

2.  A new inspection method

Despite promising results, electromagnetic NDT techniques deployed to externally inspect finned tubes were shown to be limited by access due to HRSG design. We therefore determined to develop an internal NDT technique which can be remotely deployed via headers through standard inspection stubs. The initial challenge for inspecting finned tubes in HRSGs is first to be able to deploy the NDT probe at a chosen location.

The Eddy Current (EC) technique selected was the Near Field Array (NFA) technique. In its basic form, the Near Field Technique (NFT) relies on a driver and pickup coil probe that measures the lift-off, and converts it to amplitude based signals. The receiver coil is close to the transmitter coil in order to measure the variation of the near[S1] -field zone of the driver coil. The magnetic field in the near field has very low penetration and is therefore not affected by features such as fins, hangers etc. NFT performs well on the internal surface of finned tubes since the eddy currents do not penetrate the tube wall. NFT is specifically suited to detecting corrosion and erosion inside carbon steel tubing. Recent improvements have made array technology more affordable and Near Field Arrays are now commercially available. The technology is based on the same principle as NFT. However, in this case the single pick-up coil is replaced by a ring of smaller coils acting as individual receivers. The coils behave therefore like an array allowing for C-scan imaging.

3.  Experimental apparatus and samples

The experimental data acquisition setup was composed of an NFA probe with a frequency range of 1-40 kHz and a commercial ECA instrument. The display window is composed of the following sub-windows from left to right, Absolute strip chart, Processed strip chart, C-scan, Absolute Lissajous display (top right corner) and Processed Lissajous display (Bottom right corner).

4.       Parameters: gain and frequency

To carry out the experimental work, the gain and the frequency had to be determined to optimise data quality. Those parameters were evaluated by studying their effect on the Signal to Noise Ratio (SNR) from Flat Bottom Holes (FBH) at depth relative to the wall thickness of 25, 50 and 75 percent. SNRs from two transverse notches have also been computed.

All experiments involving SNR measurement have been carried out using Sample #2 as it is more representative of a real component. The internal 3mm FBHs (D, E and F) and Notch1 and Notch2 were used. The choice was motivated by the ability to image these artificial defects consistently.

4.1 Inspection gain

The inspection gain is the master value that can be adjusted in the instrument. This gain cannot be adjusted during the data processing. From this point it will be referred to as Real Gain.

In order to study the influence of real gain all other parameters such as frequency and processing have been fixed. The common trend between the FBHs and the notches is that larger defects like FBH-75 percent and Notch2 achieve their highest SNR at low gain compared to shallower defects. This can be explained by the fact that the deeper defects represented by FBH-75 percent and Notch2, generate a clearer change of phase in the induced magnetic field. The same trend is observed with the notches. Furthermore, a second trend is observable which shows that the SNR for all defects plateaus from 54db to 60dB. This shows that there is saturation of the signal and beyond a certain gain level, the SNR does not change.

4.2 Inspection frequency

In this section, the influence of the frequency on data quality was studied. The SNR was measured for the same defects against a range of frequencies between 8kHz and 40kHz. The SNRs for the 25, 50, 75 percent FBH as well as for Notch1 and Notch2 have been measured and reported in Table 2. A setup providing scans allowing the imaging of the FBHs and Notches was established for which the real gain was set at 48dB and only the frequency was modulated from 8kHz to 40kHz. It can be seen in Figure 9 that for all defects, the highest SNRs are obtained for a frequency of 14kHz apart from defect Notch1 for which the peak SNR is reached at 8kHz. Beyond 14kHz the SNR amplitude decreases. In this case it easy to establish the frequency of 14kHz as the optimum value for the rest of the experiments compromising slightly for Notch1.

5. Results and discussion

The gain and the frequency determined via the calculation of the SNR have been used to image the samples described in section 3.

In Figure 12 the grooves are all detected. In order to build the depth sizing curve shown in Figure 13, the spread of individual grooves is evaluated using the Absolute channel. The depth related to the amplitude is extracted using the phase information in the Lissajous plot. It can be noticed that in this case the curve is slightly inconsistent which could be due to probe movement in the tube which can be adjusted according to the inspection. Table 3 shows a very close agreement between the machined depths of the groove versus the measured depth using the NFA technique.

Figure 14 and Figure 15 show respectively the artificial FAC from sample#1 and the simulated wall-loss from sample #2. Both defects are detected and the FAC spread can be evaluated. However, it should be noticed that the artificial FAC was too shallow to allow an accurate depth sizing. The taper simulating a wall loss is clearly visible in the absolute channel. The slow drifting can be evaluated using the groove sizing curve.

6. Conclusions

The Non-Destructive Technique developed was shown to be capable of detecting the most common defects encountered in HRSG finned tubes such as FAC, pitting and wall loss. Transverse cracks simulated by slots can also be detected and sized in length and depth. Following recent site trials, it is anticipated that the finished system will provide a unique NDT system providing improved detection and quantification capabilities. It is expected that a modified NFA probe will deliver a higher resolution allowing smaller pits to be detected and measured as well as axial cracks.

The authors' thanks to previous work by ASME, Doosan Babcock and Electric Power Research Institute.

About the authors: Steve Kenny has worked in the field of NDE and Integrity Management for over 40 years, and for around 25 years, has been dedicated to the life management of ageing Power Plant. He has worked for Doosan Babcock for 10 years after previously working for Utility National Power and Research Organisation TWI. His current role is Senior Business Development Manager for Asset Management of Thermal Power Plant, including Critical Component Replacement. He can be reached at steve.kenny@doosan.com.

Agnieszka Kuczynska started working for Doosan Babcock in Renfrew, UK, as a R&D engineer in the combustion team in 2008. In 2012, she took up the role of project engineer in the non-destructive testing development group under asset integrity management. She has co-authored three patents proposing novel engineering solutions in the power industry.

Damien Pain is a project engineer for Doosan Babcock. From 2006-2009, he worked as a project leader at The Welding Institute in x-ray digital radiography, x-ray computed radiography, ultrasonic testing and eddy current testing.