Implementation of Statistical Analysis of Inspection Data ... · Fabric maintenance Inspection...

Copyright © Sonomatic Ltd 2015 – Restricted Commercial not for distributionwww.vsonomatic.com

www.sonomatic.com

Implementation of Statistical Analysis of Inspection Data to

Improve Integrity Management Process

7th Middle East NDT Conference & Exhibition, Kingdom of Bahrain, September

2015

Vichaar Dimlaye, Sonomatic Ltd, Abu Dhabi

Copyright © Sonomatic Ltd 2015 – Restricted Commercial not for distribution

Integrity Management

• The main objective is to identify and implement

actions to avoid failure by leaks or rupture for

pressure equipment.

• It is a life cycle process which is applicable to all

stages of an asset’s life.

• Operations comes with many risks, thus cost effective

decisions are required for safety and reliability of such

assets.

• To make effective decisions, it is important to

understand the associated risks and this relies on

sound knowledge of the equipment condition and

how that may evolve over time.


Integrity Management Decisions

• Decisions have business impacts :

a. Repair or replacement of pressure vessels,

pipework and pipelines

b. Re-rating

c. Frequency and type of inspection

d. Detailed Fitness for Purpose studies

• These decisions are based on evaluation of

information from a range of sources and with varying

levels of uncertainty.

• It is important to understand the accuracy of the

information for reliable decision making.


• Inspection is aimed to provide direct indication of

equipment actual condition– but there are (often

significant) uncertainties associated with inspection

information.

Inputs to IM Decisions

Integrity Management

Decision Making

Corrosion Risk Assessment

Hazard Identification

Consequence Assessment

Risk Based Inspection Planning

Corrosion Control

Corrosion Monitoring

Process monitoring

Process chemistry

Flow assurance

Sand Assessment

Fabric maintenance

Inspection

Compliance and regulatory

Design and fabrication info

Effectiveness of

decisions depends on

the relevance of input

information

• Effectiveness of decisions made relies on the

accuracy of the theoretical based information and

indirect measures which comes with uncertainty.


NDT Data and Decision Making

• Inspection provides key information for use in corrosion and

integrity management.

• Inspections will always have limitations, these being

dependent on the type of NDT techniques and condition of

the equipment.

• These limitations affect the extent to which inspection

provides information on the real condition of the equipment.

Perfect knowledge of

equipment condition

Optimal decisions

Good knowledge of

equipment condition

Poor knowledge of

equipment condition

Good decisions Poor decisions

Increasing operational costs and consequential costs of failures


NDT Data and Decision Making

• Understanding the limitations of the inspection data and making best use

of the available information is fundamental to integrity management

decisions that are aligned to the real condition of equipment.

• The significance of this is often overlooked in practice.

• Failing to understanding the limitations of inspection data and on how to

make best use of the data often affects decisions. Hence, leading to major

business impacts as

follows:

• Unnecessarily short inspection intervals leading to additional

inspection cost

• Excessive expenditure on corrosion control

• Unnecessary or premature replacement involving large capital cost


Impact of Inspection Limitations on IM

Decision Making


Pipework Example

• Most operators make use of RBI. Decisions

on intervals and locations are updated on

consideration of inspection results.

Corrosion Risk Assessment Integrity Assessment

Consequence of Failure Probability of Failure

Assessment

Risk ranking

Define intervals Define locations

Inspection

Integrity review

This part of the

process is only as

good as the

inspection


Pipework Example

• 0 degree manual ultrasonic (MUT) wall

thickness measurement is the main

form of inspection for pipework (>2”) in

most process industries.

• The approach typically involves

scanning over grids at selected

locations and recording the minimum

thickness.

• Inspection conditions are challenging

involving access restrictions and

variable surface conditions.


Pipework Example

• Sonomatic has carried out statistical analysis of many large (1k’s -10k’s point) inspection history databases for UK offshore operators

• The graph illustrates typical in-service performance of MUT for pipework where there is no corrosion active.

• Graph shows variations for repeated measurements (over a 12 year period) at the same locations in a system where there is limited actual change

-4 -3 -2 -1 0 1 2 3 4

0.0005

0.001

0.0050.01

0.050.1

0.25

0.5

0.75

0.90.95

0.990.995

0.999

0.9995

Difference relative to mean value for location (mm)

Pro

babili

ty o

f short

fall

Measured

T-location

Approx 10% (<0.01) of readings have

underestimation of more than 1 mm.

Approx 10% (>0.99) of readings have

overestimation of more than 1 mm.

Measurement Errors (mm)


Pipework Example

• IM system will tend to focus on areas where the readings are lowest.

• Given the high error levels in the inspection, in any situation in which there are a large number of readings, the extremes will tend to be those measurements with the largest errors.

-4 -3 -2 -1 0 1 2 3 4

0.0005

0.001

0.0050.01

0.050.1

0.25

0.5

0.75

0.90.95

0.990.995

0.999

0.9995

Difference relative to mean value for location (mm)

Pro

babili

ty o

f short

fall

Measured

T-location

Decision making can be driven by the

measurements with the largest errors.

This is what happens in many current

database driven approaches.

Measurement Errors (mm)


Pipework Example

• Its not only current condition that is of interest. Remaining life has to be considered in decision making – rate of degradation is important.

• Impact of measurement error tends to be more significant on corrosion rate.

• Negative corrosion rate tend to be ignored-but they are indicative of error.

• Example from a system with some active degradation. -2 -1.5 -1 -0.5 0 0.5 1 1.5 2

0.0010.003

0.01 0.02

0.05

0.10

0.25

0.50

0.75

0.90

0.95

0.98 0.99

0.9970.999

Corrosion rate (mm/yr)

Pro

babili

ty o

f short

fall

Approx 35% of locations

have negative corrosion

rate – indicator of high

error levels in rate

This is where the drivers for integrity

management decisions will tend to be.

This coincides with the region in which the

largest errors can be expected.


Pipework Example

• Presence of measurement errors can give rise to inappropriate decisions being made.

• The priorities for inspection in current practise for many RBI managed systems are defined as:– Locations where the “wall loss” relative to nominal is largest,

– Locations where the “corrosion rates” are highest,

– Locations where the shortest “remaining life” is calculated.

• All three points are often locations with largest errors.

• The above are established by direct analysis of the information in the inspection database, without any understanding or knowledge of the implication of the measurement errors.


Pipework Example:

Improved Approaches to Reduce Measurement

Errors• Accuracy and reliability of thickness measurement methods can be

improved by:

• Inspection techniques, e.g. automated corrosion mapping vs MUT

• Inspection procedures, e.g. set-ups and locations for repeat inspection

• Minimisation of surface related error, e.g. paint thickness correction

• IM decisions are made on large datasets and hence need for appropriate statistical analysis to assess the impact of measurement errors.

• A range of statistical analysis methods can also be applied to assist in more reliable decision making, i.e. decisions better aligned to actual condition and variation within a system.

• Closer integration of inspection and analysis activities is essential to making best use of inspection within integrity management.


Implementation of Statistical Analysis

• Statistical analysis methods are fundamental to understanding the effects

of measurement error and making appropriate allowances in decision

making.

• Statistical analysis methods are also applicable to assessments of

corrosion. Corrosion processes across a wide range of conditions are

found to be amenable a simple parametric description. For example,

analysis of Sonomatic corrosion mapping data indicates typically

regular behaviour of the corrosion process for the following:

− Pressure equipment (vessels, pipework, pipelines) with active CO2 corrosion

− Pressure equipment with H2S corrosion

− Pressure vessels and pipework with internal O2 corrosion

− Pressure vessels with external O2 corrosion (CUI)

− Pipework with naphthenic acid corrosion

− Pipework and tanks with under-deposit corrosion

− Petrochemical and minerals refinery vessels with acid corrosion


Statistical Behaviour of Corrosion

8 9 10 11 12 13 14 15 16 17 1810

-5

10-4

10-3

10-2

10-1

100

Example - CO2 corrosion

Thickness (mm)

Pro

port

ion o

f are

a

Normal

distribution

Corrosion

• There are two distributions:

• Normal distribution associated with as-

manufactured wall thickness variations of the

material.

• Exponential distribution (linear on log

scale) representing the distribution over the

degradation area.(Approx. 10% of the

inspected area)

• Corrosion mapping data is from a

lower part of a separator vessel and it

has suffered from CO2 corrosion.

• The wall thickness distribution

generated from the corrosion mapping

data.


Statistical Analysis in Non-Intrusive

Inspection (NII) of Vessels

Inspectionmm 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200

µs

13

14

15

16

17

18

19

20

NII Assessment and definition of requirements

CO2

nx

i i

i

j j

j

P(X )P( )P( | X)

P(X )P( )

0 100 200 300 400 500 600 700 800 900 10004

5

6

7

8

9

10

11

12

13

14

Extent of corrosion (mm)

Allo

wable

wall

loss (

mm

)

Shell

Ends

Evaluation of inspection including

Statistical Analysis of resultsJustification

Workscope meeting

requirements

8 9 10 11 12 13 14 15 16 17 1810

-5

10-4

10-3

10-2

10-1

100

Example - CO2 corrosion

Thickness (mm)

Pro

port

ion o

f are

a


NII Statistical Analysis

• Statistical analysis is a fundamental requirement to NII carried out in accordance with the HOIS Recommended Practice.

• Outputs of analysis cover– Is degradation present

– Estimates for minimum thickness

– Estimates of probabilities for limiting conditions

– Evaluation of whether the inspection meets the requirements, i.e is the coverage sufficient, is the information provided by the technique suitable

• Outputs of analysis also feed into the RBI– Updated estimates of worst degradation

– Corrosion rate estimates

– Nature of distribution of degradation

• The analysis also allows assessment of the capability and limitations of the inspection technique as deployed.


Statistical Analysis for Inline Inspection (ILI)

• Current analysis for ILI data tends to ignore negative corrosion rates, which are due to measurement errors.

• Also, ignoring errors associated with positive corrosion rates.

• For illustration, let’s consider two ILI runs which were carried out on a pipeline.

• Figure represents the wall loss for the matched defects for the two ILI runs.

0 5 10 15 20 25 30 35 400

5

10

15

20

25

30

35

40

Measured depth 2009 (%)

Measure

d d

epth

2011 (

%)


Statistical Analysis for Inline Inspection (ILI)

• Normal probability plot is generated for the matched defects data,

• Red line is the normal distribution fit to the data with mean of -0.5% and a standard deviation of approx. 4.6%.

• Negative mean, suggest on average no corrosion has occurred between 2009 and 2011 for the matched defects.

• Symmetrical nature of the differences indicates most of them are due to measurement errors,

• As such, this will strongly affect any estimation of corrosion rate.

-15 -10 -5 0 5 10 150.003

0.01

0.02

0.05

0.10

0.25

0.50

0.75

0.90

0.95

0.98

0.99

0.997

Difference in measured depth (%)

Pro

bability


Correction of Measurement Errors

• Correction approach developed by Sonomatic based on the distribution of measurement error and use of an appropriate distribution for the depths of the matched defects.

• Deconvolution process is carried out on the measured data to produce the Weibull distribution (in red) and the measurement error (in green).

• Measured data shows an overestimation of the actual wall loss.

• With this approach and considering the measurement error, more realistic actual condition and hence remaining life can be derived.


Summary• Integrity management relies heavily on information on equipment

condition gathered during inspection.

• Inspection approaches have limitations affecting reliability of the information.

• As such, integrity decisions can be severely affected if these limitations are not well understood and considered.

• Improvements can be made in terms of inspection approaches to provide increased accuracy and reliability for more effective decision making.

• The use of statistical analysis ensures the value of the inspection data is maximised and decisions are aligned to provide more reliable view on actual condition of equipment and its remaining life.

• Closer integration of inspection and analysis activity is essential to making improvements to integrity management practise that Operators are beginning to expect.


Thank You!

Implementation of Statistical Analysis of Inspection Data ... · Fabric maintenance Inspection...

Documents

Transcript of Implementation of Statistical Analysis of Inspection Data ... · Fabric maintenance Inspection...