Table of Contents ECNDT '98 Session: Chemical, Petrochemical

How Reliable is Acoustic Emission (AE) Tank Testing? The Quantified Results of an AE Usergroup Correlation Study! P.J. van de Loo - Shell, The Netherlands. B. Herrmann - Dow, Germany.

Introduction

Acoustic Emission (AE) for bottoms of vertical atmospheric storage tanks is a non-intrusive technique which is more and more widely used. It is an empirical based technique with very atractive claims. The experience has mainly been built up by the contractor Physical, Acoustics Ltd.(PAL), who also formulates the claims (Ref. 1, see also section 2). Although PAL are very communicative in debating backgrounds of the technique, the actual criteria on which the gradings are based, are company proprietary and the reliability of the technique is questionable. For customers and also for authorities, who are responsible for decision making for their further inspection and maintenance programme, this is a situation which is not accepted by all parties.

To solve this the following strategy has been developed: before debating the acceptance of the technique with the authorities, the AE users from the petro-chemical industry first to convince themselves of the reliability of the technique. To this end an AE useroup collected information of a representative population of storage tanks of which both the AE-gradings had been determined, as well as their actual condition in a full off-line inspection programme. This correlation study resulted in quantified figures about the reliability of AE testing, which will be discussed in more detail in this paper.

2. Short description of the AE technique

Taken from PAL brochure (Ref. 1):

"TANKPAC is not an inspection method, but a 'sorting' system which can separate 'good' tanks from 'bad' and so direct maintenance to where it's most needed. Highly sensitive acoustic sensors are attached to the tank wall and the tank monitored, following a period of conditioning during which valves are closed and heaters/agitators turned off. The fracture of corrosion products is detected together with leaks which are active during the actual monitoring period. A percentage of this data is located by triangulation but the most important information from a maintenance management point of view is the overall condition of the floor which is given a grading on an 'A' to 'E' scale."

The ratings and indicative recommendations are:

A grade	very minor	no maintenance necessary
B grade	minor	no maintenance necessary
C grade	intermediate	some maintenance is needed
D grade	active	give priority in maintenance schedule
E grade	highly active	give highest priority in maintenance schedule

In terms of follow-up recommendations it is mentioned that the AE information should always be combined with the other (historical) information available. When no full off-line inspection or maintenance is planned, recommendations for AE retest intervals are given, varying from 5 to 1 year for increasing AE grade.

In Figure 1 (Ref. 1) for about 600 tanks, tested by PAL worldwide, the AE scores are given. Almost all of these tanks were scheduled for mandatory maintenance and/or inspection in accordance with internal, API, EEMUA or local requirements.

Figure 1. Distribution of percentage of tanks per AE grades. Total population 598 tanks, of which 129 crude oil tanks. Source: PAL, 1996 (Ref. 1). Note: Population exists almost completely out of tanks, which were scheduled for periodic inspection. Only of part of these follow-up inspection data are available.

We observe more than 50 % of these tanks have an A- or a B-grade score. If this indeed would be confirmed by findings that "no maintenance is necessary", one would be able to save a lot of money, which better can be spend on prioritising the D- and E-grade indications, which represent higher risk tanks.

On the basis of feedback of the tankowners to PAL, PAL was able to classify 80 tanks in 4 classes of follow-up condition:

FU1	No damage/no repair
FU2	minor damage/no repair
FU3	damage/some repair
FU4	damage/major repair/new floor

As the industrial AE usergroup we wanted to have in independent check on the claimed findings.

3. Description/classification of the tanks and their inspection follow-up grading

Accompanied by the necessary background information, within the usergroup an inquiry form was distributed. In Table 1 a specimen of such an inquiry form has been included. As a result of the inquiry about 70 completed data sheets were collected. From analysis these appeared to give similar results as those earlier presented by PAL. Based on these results it was found justified to combine both data sets and to bring them in one format. This allowed more comprehensive statistical analysis, of which the main results are presented in Figure 2.

Table 1. Example of tank data record sheet as incorporated in the data base (Ref.2).

Follow-up results versus AE-grades, normalised per AE-grade
Population of 148 tanks (Shell, Dow-Stade, DSM, Q8, Total and PAL-data base)

Figure 2. Correlation between AE grades and assigned follow-up grades: Percentages have been normalised to 100 % per AE grade. The percentages shown can be used to identify the number of "misses" and "false calls" for a statistical approach. Note: By making use of additional information (historic data, potential leak data from refined AE analysis) the overall correlation may be improved significantly.

Figure 3. Distribution of tanks in population which has been used for the statistical correlation study of Figure 2 (thus from all of them the follow-up grade has been determined). ( 33 crude, 115 product tanks ) Note that part of this population is also part of the population as given in Figure 1.

Some remarks have to be made:

	For a proper statistical correlation it is of utmost importance that the test population has a similar distribution as the total population of tanks that is normally subjected to AE testing (Figure 1). Figure 3 shows that this is the case.
	With regard to the determination of the follow-up grade the information is rather subjective: in the inquiry form, similar to the original classification of PAL, a distinction was made between class 1 (no damage/no repair) and class 2 (minor damage/no repair). Further inquiries gathered at the maintenance and inspection people of the refineries revealed that this distinction could not be considered as relevant: some called a little brownish corrosion appearance no damage, other called it minor damage. The main, and for maintenance the only relevant criterion, is whether one had decided to take the tank again in operation with or without any repair action. This formed the rationale to combine these two classes into one: FU 1/2: no repair required.
	With regard to class FU3 (damage/some repairs) there were also some differences in interpretation: some maintenance people classified their tank in FU3, when they repaired a couple of small pits ("We did the repair, because we had access to the cleaned tank, but we would never have opened it for this reason"), whereas others classified a real necessary repair of an almost through the wall pit , also as a minor repair, thus also in the FU3 class.

Taking these considerations into account, we judged the overall approach of a division into three follow-up classes justified. These classes represent the global condition of a tank, on the basis of which maintenance people would plan their actions. In fact "To open or not" is the most meaningful parameter that should be correlated with the AE-grades, the urgency of opening may be determined by the severity: D or E. When a tank will be opened, the degree of repair will be decided on the basis of the follow-up inspection anyhow.

4. Additional information

One must keep in mind that decisions on maintenance and inspection items never will be taken on the limited information as presented in the data sheets of Table 1, thus neither on the AE data alone, nor only on the historical information.

In the statistical correlation as presented in Figure 2, use has only been made of the overall grading, whereas in the routine AE test procedure of PAL, additional information such as "potential leak data" is always made available. "Potential leak" data are indicative for heavy corrosion spots and/or potential leaks. The value of this additional information can best be illustrated by Figure 4. Figure 4 shows an example of 2 tanks with similar overall C-grades, but distinctively different potential leak indications. At the follow-up inspection it was found that the tank H had moderate overall corrosion, whereas tank P had some severely corroded soots at the underside of the bottom plate at the position as indicated by the "potential leaks". There are also examples of tanks where minor leaks were identified, which could not or hardly be detected with an off-line inspection, when the bottom is not fluid loaded.

Figure 4. 3-D location plots of two different tanks, both rated with an overall C-grade (upper Figures), but Tank P showing localised C-and D-grades in the "potential Leak data plot (lower left Figure), whereas Tank H had no "potential leaks". The indications were confirmed by the follow-up inspection.

5. The underlying data base

All 150 data sheets have been collated in a separate report (Ref.2). Because the tank information is proprietary of the oil and chemical firms, the information is coded. The database is meant to give tank owners and authorities insight in the information that has been used in the correlation study. We hope that the data base demonstrates that the tanks that have been used are representative in terms of age, product containment, size, etc. for tanks that are normally scheduled for periodic inspection. If somebody feels the need for more detailed information on individual tank data, he may ask the authors of this report to mediate with the tank owners.

6. Examples and Remarks

The database as presented here must be considered as a snapshot. The usergroup consisted of a limited number of users, not all existing information was collated and since the establishment of the correlation study more useful data have become available. There are no indications that the later findings deviate from the overall trend as presented here. Although one may feel it very useful to regularly update this database, which can be done by giving feed back information on follow-up inspections to PAL, it is felt by the aut ors that more can be gained by paying attention to more dedicated data analysis algorithms, than to extending the overall correlation. Because it has clearly been demonstrated that e.g. the application of the filtering algorithm for the "potential leak data" results in more detailed information which supports decision making on individual tanks (see Figure 4).

An important criterion for the judgment of the reliability of a test method is information about the repeatability of the method. Figure 5 shows the results of tests performed on one tank in 3 consecutive years. The overall damage patterns appear to be well repeatable, the overall grades were C, D and C respectively, thus not fully consistent. The same holds for the "potential leak" data.

Figure 5. Repeatability of AE tests, performed on one tank in 1995, 1996 and 1997. Above: the overall data, being C-, D- and C-grade respectively. Below: the "potential leak" data, having "none, C- and D-grades, and A- and B-grades respectively.	Figure 6. Result of the follow-up inspection of the tank of Figure 4 with Rosen's TBIT Magnetic Flux Leakage (MFL) bottom scanner. Indicated are all defects with a wall thickness reduction of more than 40 %.

Finally we want to show one striking example of the correlation between AE data, obtained on a single crude oil tank and the follow-up inspection data as obtained by Magnetic Flux Leakage (MFL) bottom scanning, performed by Rosen with their TBIT (Ref. 3). The TBIT software allows us to put any threshold at the presentation of the registered defects. In Figure 6 only defects with a registered wall thickness reduction of more than 40 % of the nominal plate thickness of 8 mm, have been presented. We observe a remarkable resemblance in the configuration of the defects..

The information presented above indicates that AE testing always should be considered as a global test and that decision taking is not black and white "yes or no". In case of doubt as mainly with the B-, C- and D-grades, one can decide to postpone maintenance but define an AE retest interval of e.g. 5, 3 or 1 years. Then decisions may be taken on the basis of trending, whereas also a better maintenance planning within the operational constraints is made possible.

7. Conclusions

The performed correlation study has resulted for the members of the AE usergroup in confirmed/increased confidence in the AE testing technique for atmospheric storage tanks.

The statistical correlation between AE grades and inspection follow-up grades has revealed quantified figures on the reliability of AE testing.

Most strikingly is the demonstrated 100% correlation between A-grades and confirmation by follow-up inspection that no repair was required.

Increasing AE severity indications correlate well with more damage and larger repair requirements.

By making use of information, additional to the overall AE grade, such as the tank historie the "potential leak" data from refined AE analysis, the overall correlation will IN-improved significantly.

"Misses" in the AE-grade indications may be coped with be proper consideration of retest intervals.

References

1. Physical Acoustics Group, "Acoustic Emission Testing of pressure systems and tank", Brochure PAL, Over/Cambridge, UK
2. P.J. van de Loo and S.N. Gautrey, "Correlation of AE gradings and their inspection follow-up gradings: report containing the underlying data base", to be issued and available on request at the authors, Shell Amsterdam / PAL, Over / Cambridge.
3. Brochure on TBIT tank testing of Rosen Inspection Technologies, Zoetermeer. The Netherlands

Acknowledgment

The authors would like to acknowledge the maintenance and inspection People from all refinery and chemical plants from Shell, Dow, Exxon, ICI, DSM, Q8, Total, who provided all the required feed back information. We would like to give special acknowledgment to PAL, who initiated to feed back data base approach, which enabled this study and their openness to debate the various aspects of AE testing.