Automatic Sizing of Internal Pits in Heat Exchanger Tubing

by Monty O’Connor*

* Eddy Current Technology, Inc., 201-A Horace Avenue, Virginia Beach, VA 23462; (757) 490-1814; fax (757) 490-2778; email: montyect@gmail.com.

Abstract

When internal pitting occurs in heat exchanger tubes, there are often thousands of pits per tube, making completely automated detection and sizing of these pits highly advantageous.

The best method for sizing I.D. pits in 90-10 copper nickel tubes according to EPRI (Electric Power Research Institute) research is to use the vertical volts method. As the occasional external defect is encountered, this would result in erroneous readings if these signals were analyzed by using vertical volts in order to calculate the depth of an internal pit. To apply the vertical volts method for internal pit sizing with the ect MAD 8D software, it was necessary to first measure defect depth by phase angle to make a determination as to whether the signal originated from an internal or external surface. If phase indicated the pit was on the internal surface, then the vertical volts calibration curve was applied to measure the internal pit. (Citation) Electromagnetic NDE Guide for Balance-of-Plant Heat Exchangers, Revision 2, Sections 3.1 and 5.1.

The result was a software package which without dependency upon the user to find and record the depth of the deepest internal pits in tubes. The software has been tested in an EPRI round robin and was recommended for automatic detection and sizing of internal pits.

Introduction

Over the past two years, Kenji Krzywosz at EPRI, has been researching eddy current measurement techniques for the accurate sizing of pits on the inside surface of heat exchanger tubes. As part of this study, naturally occurring pits were studied in 90-10 copper nickel tubes that had been removed from service. The conclusion of the study was that in this case, the most accurate method of sizing these internal pits is to measure the vertical volts of the eddy current signal using a relatively low inspection frequency.

The tube studied was a 5/8″ diameter by .049 wall 90-10 copper nickel tube. A primary test frequency of near 50 kHz may have normally been chosen, but in this study the frequency with the best results was 10 kHz.

B & W | Color
Figure 1. Auto Analysis by Phase

A tube with an internal corrosion problem could easily have thousands of pits along its length; therefore, there is a need to quickly, accurately, and reliably find the deepest pit or pits in a tube without having to rely on a manual technique. The ect Auto Analysis Software has had this capability for years (See Figure 1 and note the analysis of 83°, 1.46 Volts, 60% of Wall O.D. indicated in the sixth line of the Menu), but did not have the ability to analyze from vertical volts. We at Eddy Current Technology decided we should add this capability to our software for the ect MAD 8D eddy current system.

B & W | Color
Figure 2. Auto Analysis of an I.D. Pit Using a
Vertical Volts Calibration Curve

Auto Analysis Using a Vertical Volts Calibration Curve

To auto analyze internal pits by vertical volts, two changes had to be made to the software. The first was to add the vertical volts capability as an analysis method. This was quite easy to accomplish since the software already supported peak-to-peak measurement; so it was a simple matter to allow the selection of a new parameter for measurement, and then to analyze based on only the vertical portion of the signal (See Figure 2 and note the analysis of 25°, 3.81 Volts, 50% of Wall I.D. indicated in the sixth line of the menu).

B & W | Color
Figure 3. Auto Analysis of O.D. Defects Using a Vertical Volts
Calibration Curve Results in Obvious Errors for O.D. Defects

You cannot use a vertical volts to analyze all defects in a tube because when you use the vertical volts calibration curve to analyze an outside defect, the resulting analysis is very incorrect (See Figure 3 and note the analysis of 82°, 1.41 Volts, 45% of Wall I.D. indicated in the sixth line of the Menu). In this case, the 60% O.D. pit in an ASME calibration tube has been analyzed as 45% I.D.

Figure 4

Figure 4. Phase Analysis for O.D. and Vertical Volts Analysis for I.D.

This is resolved by initially analyzing defects using a phase calibration curve. Based on this, if the angle of the indication is greater than or equal to that of the through wall hole in a calibration tube, then the signal is from an outside defect and is analyzed with the phase curve. If the angle of a defect falls between that of liftoff (Horizontal) and the through wall hole, then the signal is considered to be an internal pit and is analyzed by a vertical volts curve (See Figure 4).

Dealing with Support Plates

Internal pits will occur at or near support plates, making it necessary to deal with support plate signals. A 40-10 kHz mix does a good job of eliminating the support plate signals; however, two factors make the support plate signal problem more difficult to deal with than normal.

First, the vertical volts method for sizing internal pits as developed by Kenji Krzywosz at EPRI, indicated that best results were obtained by using relatively low frequencies, specifically 10 kHz in this case. This frequency is lower than what might have otherwise been chosen for analysis by phase. This lower frequency penetrates the tube wall more easily, resulting in larger signals from support plates, making it more difficult to cancel the larger signals.

A second factor which occurs in the round robin testing discussed below, the mockup heat exchanger used in this round robin has non-ferromagnetic support plates. Non-ferromagnetic support plates give substantially larger support plate signals than would be obtained from ferromagnetic support plates, making it even more difficult to cancel the larger signals.

The two factors listed above result in relatively small defect signals in the mix channel. This problem is overcome by using the Mixer Gain Control to amplify the mixer signals to a level that can be worked with conveniently.

Round Robin Testing

Having made these software changes, it was then time to test the software on some real tubes.

EPRI had prepared a mockup heat exchanger which involved seven tubes: a calibration tube and six tubes which had been removed from service with internal pits that formed while in service. The capability of the software was to be tested by measuring its ability to detect all pits down to 20 per cent through wall. To avoid missing defects that we may have under called by a few percentage points, we were asked to report on all defects down to 15 per cent of wall.

For the inspection of the mockup heat exchanger, a probe drive with a speed of two feet per second was used for collecting the data. Then the data was analyzed by the Analysis Software. At the analysis stage, the software appeared to reliably analyze all defects in excess of 30 per cent of wall and all defects down to 20 per cent through wall provided that the defects were not near a support plate. Defects less than 30 per cent of wall that were close to or underneath a support plate required Operator intervention to ensure an accurate analysis. The problem was that the residual support plate signal in the Mix Channel moved the end points of the defect signal clockwise an amount that resulted in the Data Analysis Software concluding that the defect was on the outside of the tube. The Operator had to move the cursor slightly in order to force analysis by the vertical volts calibration curve. For these small defects at the support plates, it is these numbers that were used in the final report.

Round Robin Results

The large signals from non-ferromagnetic support plates at the low frequencies used for the vertical volts method of sizing internal pits played a significant roll in the result.

Requirement: Measure all pits over 20% of wall.

Tube: 90-10 Copper Nickel removed from river water service.

Results:

Auto detected and analyzed with the required accuracy:

1. All defects greater than 28%, regardless of location.

2. All defects not near support plates.

3. Possibly missed defects 28% or less that were near support plates. (Note 1)

Away from support plates all the required defects (all defects over 20 per cent) were detected and analyzed with the required accuracy to meet the EPRI guidelines.

At support plates, all defects in excess of 28 per cent were detected and analyzed with sufficient accuracy to meet the requirements of the EPRI guidelines.

There were a total of five participants in this round robin. Four used operator dependent manual analysis. Three of the four operator dependent manual analysis participants failed to meet the flaw detection criteria and failed to meet the sizing criteria. One participant used the ect MAD 8D eddy current system with automatic analysis using a vertical volts calibration curve, and this system exceeded the flaw detection and sizing criteria established in the EPRI Guidelines.

The overall result is that using the ect Auto Analysis Software for a vertical volts analysis of internal pits is a recommended procedure now in the EPRI Guidelines.

Note 1: It is actually believed that all of the defects over 20 per cent and close to support plates were correctly measured, but as the probe drive used did not have a position encoder, the exact position of the defect was not accurately reported; therefore, there is some uncertainty that the correct defect was measured for shallow defects near support plates.

Field Trials

The ect MAD 8D eddy current system was taken to the field for data acquisition and data analysis using the Analysis Software on 5/8 inch diameter 90-10 copper nickel tubes in service in a K C heater known to have an internal pitting problem.

Data was acquired from 28 tubes and analyzed, and then compared to the results of the traditional manual inspection. Only the three deepest defects encountered in a tube with the automatic analysis software was reported. Less significant defects were not to be reported. This data was compared to the data from the conventional manual analysis in which only the deepest defect encountered was reported. In all cases, the defect reported by manual analysis had been detected and analyzed correctly by the Data Analysis Software accurately.

Conclusion

The ect MAD 8D eddy current system with ect Auto Analysis Software has been shown to analyze eddy current tube data reliably with no missed calls of defects over 20% of wall loss away from support plates and all defects over 28% of wall loss at support plates. This software is a reliable, accurate, fast, Operator-independent method for analyzing eddy current data.

References

1. Kenji Krzywosz, “Eddy Current Pit Sizing: Revisited”, Paper No. Krzywosz-23:1, presented at the Third EPRI Balance-of-Plant Heat Exchanger NDE Workshop, Myrtle Beach, South Carolina, USA (June 6-8, 1994). [conference paper]