Professional Documents
Culture Documents
IPC2022
September 26 – September 30, 2022, Calgary, Alberta, Canada
IPC2022-87166
Shenwei Zhang, Terry Huang, Colin Dooley, Roger Lai, Brett Conrad
TC Energy Corporation
Calgary, Alberta, Canada
Predicted SSI (cycles/yr)
12000
which is illustrated in Figure 5.
A total of 132 intermediate valve stations with immediate 10000
For a given intermediate valve station, the SSI calculated from 2000
Rainflow counting using the pressure spectrum is considered as
the actual SSI, which is compared with the predicted SSI given 0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
by the pressure decay models based on the SSI values of the Actual SSI (cycles/yr)
14000
Predicted SSI (cycles/yr)
12000
10000
8000
6000
4000
2000
0
Figure 5: Illustration of geospatial locations of pump 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
stations and intermediate valve station Actual SSI (cycles/yr)
12000
CEPA model generates predictions that are on average closer to
the unity line and have a tighter distribution along the unity 10000
further away from the unity line and has the largest scatter 4000
0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Actual SSI (cycles/yr)
12000 Interpolation
Distribution Lognormal
10000 Frechet Frechet Frechet
type (3P)
8000 =0.68275
Parameters =4.2383 =4.7544 =4.233
=0.0077
6000 in the PDF =1.3815 =1.3016 =1.3803
=0.8341
4000
Mean 1.671 1.537 2.101 1.670
2000
COV 35% 32% 44% 35%
0
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 The probability density function (PDF) of model error
Actual SSI (cycles/yr)
associated with the API RP1176, CEPA, SSI Linear
(d) Pressure Range Linear Interpolation model Interpolation and Pressure Range Linear Interpolation models is
shown in Figure 7(a), (b), (c) and (d), respectively. For clarity,
Figure 6: Unity plot of SSI Comparison the PDF and cumulative distribution function (CDF), denoted
by f(x) and F(x), respectively, are also shown on the graph with
To quantify the prediction accuracy and precision of the
their respective distribution parameters shown in Table 2.
models, the model error was then calculated. The probabilistic
These distributions can be applied to the probabilistic fitness-
characteristics of the model error, including mean value,
for-service (FFS) assessment program for explicitly accounting
coefficient of variation (COV) and best-fit distribution type,
for the uncertainties of model error. For example, FFS
associated with each of the four models were derived using the
assessment of dents or cracks on liquid pipelines using
132 data points shown in Figure 6 and are shown in Table 2. In
probabilistic assessment method.
addition, the values of the parameters involved in the
probability density function (PDF) of the distribution are also
include in Table 2 for facilitating the application of these
distributions. Results as shown in Table 2 can also be used to
quantitatively compare the accuracy and precision of the four
models. The following observations are made:
Firstly, the CEPA model among the four models considered
has the lowest mean value (or bias) of 1.537 and the lowest
COV value of 32%, which indicates that the CEPA model is the
most accurate and precise model in predicting the SSI
compared with the other three models.
Secondly, both the mean and COV values associated with
the API RP1176 model and Pressure Range Linear Interpolation
model have negligible differences. This demonstrates that the (a) API RP1176 model
elevation does not affect pressure change and can be ignored
for simplicity.
Lastly, SSI Linear Interpolation model has the highest
mean and COV values indicating this model has the poorest
accuracy and precision among the four models. This suggests
that a non-linear decay model characterizes the SSI better than
a linear decay model. For example, the CEPA model described
in Section 3.2 explicitly uses a non-linear function to
characterize the decay of SSI and the other two models
described in Sections 3.1 and 3.4 assume the pressure range
follows a linear decay function which implies SSI follows a
non-linear decay function considering the relationship between
SSI and pressure range represented by the Miner’s rule. (b) CEPA model
Assessment
D, t, grade, dent depth
3D ILI geometric data
Pressure range bins (maximum
pressure, minimum pressure,
Level 2 pressure range, number of cycles
for each range)
Pressure @ ILI
Shape parameter-dependent S-N
curve
(d) Pressure Range Linear Interpolation model D, t, grade (strain-stress curve)
3D ILI geometric data
Figure 7: PDF of model error Level 3
Pressure range and its
5. APPLICATION corresponding number of cycles
This section presents the process of how to apply the In Table 3, each fatigue assessment methodology may
pressure decay models in the FFS assessment. The Paris law is require different input of pressure characteristics as indicated
commonly used for crack fatigue assessment, which requires by the italicized texts. Regardless of the pressure input for
the input of a pre-determined constant pressure range and its fatigue assessment of crack or dent, the process considering the
corresponding equivalent number of cycles on an annual basis. location specific pressure cycling to evaluate the fatigue life of
For dent fatigue assessment, API RP1183 recommends a given pipeline is described as follows:
different screening and assessment methodologies as shown in Step 1: identify the upstream and downstream pump
Table 3, where D and t denote the pipe outside diameter and stations for a given pipeline section and determine the
wall thickness, respectively. geospatial data associated with each pump station;
Table 3: Summary of dent fatigue assessment methodologies Step 2: collect the discharge pressure data from the
recommended in API RP1183 upstream pump station and suction pressure data from the
downstream pump station with a predetermined acquisition
Type Method Input frequency (e.g., one minute) and a duration period (e.g., one
Dent depth year);
SSI SSI @ annual 13ksi (or 90MPa) Step 3: run the Rainflow counting analysis to calculate
hoop stress cycles the SSID and SSIS corresponding to predetermined pressure
D/t ratio ranges. Typically, the pressure range is determined based on a
Level 0 for hoop stress range of 90MPa or 13ksi and the pipe attributes at
Restrained condition
Screening