Lifetime prediction methods and

accelerated testing

Dr Sue Burnay
John Knott Associates Ltd., UK

Sue.burnay@btinternet.com

Why do we need predictive modelling?

Qualification of components aims to demonstrate that they
can continue to fulfil their safety function for their
qualified life
Requirement to simulate the ageing that would occur in the
plant, using worst case environmental conditions
Thermal and radiation ageing needs to be simulated in an
accelerated time frame but is this accelerated ageing an
accurate simulation?
Predictive modelling can be used to optimise replacement
schedules for components e.g. seals, using actual
environmental conditions

John Knott Associates Ltd 2

 What are the issues for lifetime
prediction of polymers in NPPs?
1.00E+03

1.00E+02
Temperature has little
effect in this region –
radiation dominated
DED (kGy)

1.00E+01

20 C

40 C
1.00E+00
Tending towards a constant 60 C

time to failure – thermally 80 C

dominated region
1.00E-01
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00
Dose rate (Gy/s)

John Knott Associates Ltd 3

Lifetime prediction methods

Thermal ageing – Arrhenius and its limitations
Radiation ageing – power law extrapolation
Combined thermal & radiation ageing – two
approaches to modelling
Superposition of time dependent data
Superposition of dose to equivalent damage (DED) data
Limitations of prediction methods

John Knott Associates Ltd 4

 Accelerated thermal ageing
Usually based on Arrhenius relationship, using
following equation
t1 = t2 exp [Ea/R (1/T1 - 1/T2)]
Where t1 is ageing time required at a temperature T1 to simulate a service
life of t2 at a temperature T2. Ea is the activation energy for thermal ageing
and R is the gas constant
Assumes that
A single degradation mechanism is in operation
Degradation mechanisms are the same at the higher
temperature used for accelerated ageing
Activation energy Ea is constant and value is known for
specific cable formulation

John Knott Associates Ltd 5

What to look for

Are the curve shapes the same for each temperature?
Can check by determining time to reach different levels of
degradation and comparing slope of 1/T plots
If not, degradation mechanisms are not the same at different
temperatures
Is it possible to interpolate to the required end-point?
If not, use superposition to maximise use of data and use shift
factors plotted against 1/T to determine Ea
Ea determined from slope since
Ea = R loge (t1/t2) / (1/T1 – 1/T2)

John Knott Associates Ltd 6

 Superposition principle

Often used in thermal ageing to determine activation energy

Data obtained at different temperatures are superposed by applying a
multiplying factor – shift factors a(T)
Generates a ‘master curve’ at a reference temperature Tref

John Knott Associates Ltd 7

Predicting radiation ageing – power law

extrapolation method

Dose to equivalent damage (DED) given by

DED = C. Dx
where D is the dose rate, exponent x <1, C is a constant for
the material
Simple model, applicable particularly to polyolefin
materials at near ambient temperature
Described in detail in IEC1244-2
Currently the only model available for polymers
with a reverse temperature effect
John Knott Associates Ltd 8
 Power law – generating data required

Ageing at constant temperature, several dose rates

Test data (e.g. EAB, set) plotted against radiation dose
DED values extracted for different dose rates
Log/log plot of DED versus dose rate is linear
John Knott Associates Ltd 9

Example of power law method applied

to a cable insulation material (XLPE)
Dose rate dependence of dose to 100% elongation
Dose to reach 100% elongation
for XLPE insulationfor XLPE cable
at 20C
10000 insulation, radiation aged at 25 C
log D ose to 100% elongation (kGy)

1000

100
1 10 100 1000 10000

log Dose rate (Gy/hr)

John Knott Associates Ltd 10

 Limitations of power law predictive
model

1.00E+03

1.00E+02

DED (kGy)
1.00E+01

Limited to extrapolations at 20 C

the test temperature used 1.00E+00

40 C

60 C

Extrapolation range is 80 C

limited, particularly at higher

temperatures
1.00E-01
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00
Dose rate (Gy/s)

John Knott Associates Ltd 11

Predicting combined thermal/radiation ageing

– superposition of time-dependent data

Requires matrix of test data at different dose rates

and temperatures
Time-dependent data are superposed to give
master curve using shift factors a(T,D)
Semi-empirical model links a(T,D) values to
temperature and dose rate
Described in detail in IEC1244-2

John Knott Associates Ltd 12

 Superposition of combined
radiation/thermal ageing data

Same principle of superposition is used to generate a

‘master curve’ for all of the data
Shift factors a(T,D) are a function of both temperature and
dose rate
John Knott Associates Ltd 13

Typical master curve generated by

superposition

John Knott Associates Ltd 14

 Modelling the shift factors a(T,D)

Semi-empirical equation
x
a(T,D ) = exp { −E/R (1/T − 1/Tref)} [1 + k . D . exp {Ex/R (1/T − 1/Tref)}]

Simplifies to the Arrhenius equation for D=0

a(T,0) = exp {−E/R (1/T − 1/Tref)}
At the reference temperature T=Tref, simplifies to
a(Tref, D) = 1 + k .D x

John Knott Associates Ltd 15

Fitting experimental values of a(T,D) to

the equation - example

John Knott Associates Ltd 16

 Example of superposition model used for
radiation/thermal ageing of a seal – long term
validation test at 60 C, 6 Gy/hr
SILICON E (X625R)
60 o C in air
100
B9 calculated
90
B9 measured

80

70
Compression set %

60

50

40

30

20

10

0
3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0
log time (hours)

John Knott Associates Ltd 17

Limitations and use of model based on

superposition of time dependent data

Requires a large matrix of data

Superposition only possible if single degradation
mechanism is dominant, so that curve shape
remains same
Demonstrated for wide range of polymers
(particularly seal materials)
Works best for amorphous materials, not usually
good for polyolefins or semi-crystalline EPRs

John Knott Associates Ltd 18

 Curve shape must be same at different dose rates
and temperatures to use superposition – model
should not be used for this material

John Knott Associates Ltd 19

Predicting radiation ageing – superposition of

DED data

Requires matrix of test data at different dose

rates and temperatures
Also uses superposition but here DED data are
superposed using shift factors a(T)
Shift factors often related to temperature by
Arrhenius equation
Described in detail in IEC1244-2

John Knott Associates Ltd 20

 Basic principle of method

DED data at different dose rates are superposed using shift factors a(T)
that only depend on temperature
DED values will tend towards the thermal ageing limit at low dose
rates
John Knott Associates Ltd 21

Superposition of DED data - example

John Knott Associates Ltd 22

 Limitations and use of model based on
superposition of DED data

Requires large matrix of data

Cannot be applied to materials that show reverse
temperature effects in the temperature range of
interest
Demonstrated on a range of polymers, particularly
cable materials

John Knott Associates Ltd 23

Superposition of DED data – reverse

temperature effects
Superposition of
DED data for a
XLPO cable
insulation material
□ - radiation ageing
data at ≥60ºC
41■ and 22■ refer to
data at 41ºC and
22ºC respectively
Superposition would
seriously
underestimate
degradation at low
temperatures

John Knott Associates Ltd 24

 Superposition of DED data – example where ageing
temperatures above crystalline melting point

John Knott Associates Ltd 25

Limitations of predictive models

Semi-crystalline polymers
Tend to show reverse temperature effect, with
degradation being greater at ambient temperature than
elevated temperature
Arises from recrystallisation and recombination of
radicals at higher temperatures
Power law method is currently only practical approach

John Knott Associates Ltd 26

 Lifetime prediction methods

Recommended for further reading

IEC Technical Report IEC1244-2 (International
Electrotechnical Commission, Geneva)
“Determination of long term radiation ageing in air. Part
2: Procedures for predicting ageing at low dose rates”
Details of practical methods for lifetime prediction and
their limitations

John Knott Associates Ltd 27

Predictive modelling - summary

For all predictive methods, essential to understand
underlying assumptions and limitations
Thermal ageing – Arrhenius equation of practical
use
Radiation ageing – power law method useful,
particularly for materials (such as polyolefins) that
show reverse temperature effect
Combined thermal/radiation ageing – both
superposition models have proved useful but do
require large sets of test data
John Knott Associates Ltd 28
John Knott Associates Ltd 29

Developing a test matrix for accelerated

ageing tests
Identify aims of the test - for example
Simulation of service life
Simulation of accident conditions
Predictive modelling
Generation of CM correlation curves
Identify timescales needed for the test
Identify type of data needed and number of test
conditions required

John Knott Associates Ltd 30

 Timescales of tests

Factors to be considered:
Need to minimise acceleration factors
Limit extrapolation required (particularly for thermal
ageing)
Limit maximum temperature used to avoid non-
representative degradation mechanisms
Limit maximum dose rate to avoid heterogeneous
oxidation

John Knott Associates Ltd 31

Types of data needed

Proof tests and functional tests require limited

number of test conditions
Predictive modelling requires a large matrix of
data with multiple temperature/dose rate
conditions
Generation of CM correlation curves require an
intermediate no. of test conditions

John Knott Associates Ltd 32

 Test methods for specific materials

Seal materials
Compression set used as basic indicator of degradation
Measurement of leakage rate or sealing force may be
needed for functional tests
Cable insulation and jacket materials
Elongation at break usually used as basic indicator of
degradation
CM methods available for most materials
Coatings
Adhesion tests

John Knott Associates Ltd 33

Thermal ageing tests

Factors to be considered –
If activation energy E is not known, need a minimum of
3 test temperatures
Maximum temperature must be limited – changes in
degradation mechanisms at high temperature
Test temperature should be near to service conditions
(extrapolations should ideally be <25 C)
Heterogeneous oxidation likely at higher test
temperatures and in large diameter seals or cables
Good oxygen access needed
For seals, ageing must be carried out in compressed
state representative of seal housing
John Knott Associates Ltd 34
 Radiation ageing tests

Factors to be considered –
If test matrix is for predictive modelling, need a
minimum of 3 dose rates
Simulation of accident conditions can use high dose
rates, but for service conditions use as low a dose rate as
possible
Heterogeneous oxidation will occur at high dose rates or
with thick samples (a major concern when testing seals
or whole cables)
Good oxygen access needed
For seals, ageing must be carried out in compressed
state representative of seal housing
John Knott Associates Ltd 35

Functional tests on seals

Factors to be considered –
High temperature tests
Can transient be simplified to isothermal steps?
Allow for outgassing at temperature in leakage
measurements
Low temperature tests
Is there a glass transition near minimum
temperature?
Use dry gas when measuring leakage rates

John Knott Associates Ltd 36

 Typical test matrices
Tes t type Radiation dos e rates Temperature
Predictive modelling 25, 50, 100, 200 Gy/hr 20 C
100 Gy/hr 40, 60 C
0 Gy/hr 80, 95, 110 C
CM correlation curves 50, 200 Gy/hr 20 C
100 Gy/hr 40 C
0 Gy/hr 100 C
Proof tes ts 200 Gy/hr 20 C
plus 0 Gy/hr 110 C

John Knott Associates Ltd 37

Summary
There are suitable predictive models for most
types of polymers
Modelling requires large matrix of data – may not
be feasible for polymeric components in older
plant
Design of accelerated ageing tests must consider
several factors to ensure that test is appropriate
For new NPPs, recommend that data for predictive
modelling is generated early in the build process

John Knott Associates Ltd 38