Methodology to Calibrate Fragility Curves Using Limited Real-

World Data
This paper was downloaded from TechRxiv (https://www.techrxiv.org).

LICENSE

CC BY 4.0

SUBMISSION DATE / POSTED DATE

20-05-2022 / 24-05-2022

CITATION

Bereta dos Reis, Fernando; Royer, Patrick; Chalishazar, Vishvas Hiren; Davis, Sarah; Elizondo, Marcelo;
Dagle, Jeffery; et al. (2022): Methodology to Calibrate Fragility Curves Using Limited Real-World Data.
TechRxiv. Preprint. https://doi.org/10.36227/techrxiv.19806310.v1

DOI

10.36227/techrxiv.19806310.v1
 Methodology to Calibrate Fragility Curves Using
Limited Real-World Data

Fernando Bereta dos Reis∗ , Patrick Royer∗ , Vishvas Hiren Chalishazar∗ , Sarah Davis∗ , Marcelo Elizondo∗ ,
Jeffery Dagle∗ , Alexandre Nassif† , Andrija Sadikovic† , Elli Ntakou† , Olga Soto† ,Shay Bahramirad†
∗ Energy & Environment Directorate, Pacific Northwest National Laboratory, Richland, WA 99354, USA
† LUMA Energy L.L.C., San Juan, PR 00936, USA

Abstract
With increasing human dependence on electricity and increasing energy demand, electrical infrastructure has
emerged to be one of the most critical services. On the other hand, climate-change driven extreme events progressively
increase in frequency and intensity. This is one of the main reasons for making the power grid more resilient to
extreme events during which uninterrupted power supply is crucial in keeping the consequences of the extreme event
limited. The first step towards making the power grid more resilient is to evaluate the probabilities of failure for
all assets at risk from the extreme events. Thus, this paper presents the a methodology to calibrate fragility curves.
The strength of the proposed approach is its ability to calibrate the fragility curves utilizing limited data, which
is the most common constraint in carrying out such analyses. This paper describes the calibration of transmission
tower fragility curves for Puerto Rico utilizing only damage reports for hurricane Maria. That, combined with Puerto
Rico’s wind modeling & geographic information system, this study also calibrates fragility curves for transmission
towers made with four different construction materials.
Index Terms
Resilience, Fragility Curves, Transmission, Extreme Events, Hurricanes

I. I NTRODUCTION
Designing a resilient power system has become imperative in the face of increasing intensity and frequency of
the extreme events. Every element of the power system, such as transmission and distribution support structures,
transmission and distribution conductors, substation equipment, centralized and distributed generators, and even
loads are vulnerable to a variety of hazards. To increase the overall resilience of the entire grid, it is important to:
first identify these vulnerabilities; second evaluate the probability of failures (fragility) over a range of excitation
measures stemming from the type of extreme event and; third mitigate asset fragility making relevant system
assets robust thus reducing their probability of failure (hardening the assets). The fundamentals of asset fragility,
vulnerability and risk is described in detail by Keith Porter in [1].
A summary presented in [2] lists the different high impact low probability (HILP) hazards threatening the power
system are Seismic, Hurricane, Tornado, Geomagnetic storm, High-altitude electromagnetic pulse, ice storms, cyber-
attack, physical attack, etc. While planning and preparing for such HILP events, the most cost-effective strategy
is to upgrade/retrofit the assets which are most vulnerable as well as most consequential when lost/damaged [3].
Fragility curves are used to assess these most vulnerable assets and also to compare the effectiveness of different
retrofitting strategies for these assets. For a single asset experiencing a single hazard there can be multiple causes
of failure and thus multiple fragility curves. For different causes of failure the restoration time for that asset can
be significantly different. Thus, fragility curves can also inform the restoration time calculation to facilitate the
evaluation of overall consequences to the power grid. The studies evaluating power system consequences (such as
The authors would like thank the Federal Emergency Management Agency (FEMA) and the U.S. Department of Energy (DOE) for funding
this study. This study was conducted at Pacific Northwest National Laboratory, which is operated for the U.S. Department of Energy by
Battelle Memorial Institute under Contract DE-AC05-76RL01830.
Corresponding author: [email protected]
the loss of load, energy not served, etc.) for different HILP events: 1) Augment power system models (or use more
detailed node-breaker version of the models), so that asset fragilities can be translated to power system contingencies
[4], and 2) Utilize asset fragility curves that are either developed by modeling structural integrity of assets using
tools like Structural Analysis Program (SAP2000) [3], or developed using real world data based academic models
such as log-normal functions [5]. However, the availability of real world data for infrastructure damage and fragility
is often very limited for such HILP extreme events.
The fragility curves for a hurricane hazard for different types of generation given the hazard’s peak gust wind
speed are presented and utilized in [6] and for transmission electrical conductors relative to: a) sustained wind speed,
b) relation to the conductor type, and c) the distance of support structures is presented in [7]. To that end, this paper
presents the methodology to calibrate fragility curves. The strength of the proposed approach is its ability to calibrate
the fragility curves utilizing limited data, which is the most common constraint in carrying out such analyses. This
paper describes the calibration of transmission tower fragility curves for Puerto Rico utilizing only damage reports
for hurricane Maria. That combined with the Puerto Rico’s wind modeling & geographic information system, this
study also calibrates four transmission tower fragility curves based on their construction material. The remainder of
this paper is organized as follows: Section II presents the overview of tower fragility curves for wind. Section III
describes how the maximum wind speed for every tower from historical hurricane events is estimated. Section IV
presents the proposed calibration of fragility curves. Section V describes the resulting calibrated transmission towers
fragility curves. Section VI provides concluding remarks on the proposed calibration of transmission tower fragility
curves for gust wind speed.
II. T RANSMISSION T OWER F RAGILITY C URVE
There are two main hazards for power systems in hurricane events are wind and flood. Different wind speed
measurements are used for different elements/assets. The fragility of conductors of the transmission line is commonly
not utilized as stated by the authors that have developed fragility curves of conductor [7]. Reference [8] states that
due to design requirements (i.e., civil engineering design requirements) the fragility of the transmission line under
wind loading is mainly determined by the failures of support structures (i.e., towers). Flood hazard for power
systems refers to the inundation – defined as the depth of water above local land elevation. Other parameters
associated with flood are rainfall, speed of the water (i.e., increase in water speed on rivers), and debris on the
water but those are commonly utilized for bridges and other structures. The combined utilization of wind and
flooding for transmission and distribution lines is presented in [9]. However, the combination of hazards is not
commonly utilized. In summary, substations are mainly subject to inundation depth and transmission towers to
wind speed. There are four approaches for obtaining the fragility curves [10].
1) Judgmental method – It is not limited by data or models, being a fast method to implement. Utilizing expert
knowledge to determine the fragility. Thus, it is difficult to validate and is based on the biases or past
experience of the experts.
2) Empirical method – It utilizes statistical data from controlled experiments or from actual extreme events.
It is useful and flexible if data is available. The availability of data can be scarce and experiments can be
expensive.
3) Analytical method – It is based on physical models that can be validated and verified, enhancing transparency.
It is easier to extrapolate results to new situations. It facilitates a distinction between aleatory and epistemic
uncertainty and is based on simplifications and assumptions. Require the availability of data for the models.
4) Hybrid method – It is utilized to overcome the limitations of any individual approach but have the same
disadvantages of the combination of approaches. A common combination is the Empirical and Analytical
methods, thus, the analytical model is fitted based on data to overcome the differences from the model to
actual events.
The designs of transmission lines have a well-founded desire to exercise some control over the sequence of failure
for the different line components [11]. Fig. 1 presents the different line components. The cables are especially
important under extreme events given that failure puts critical demands on the failure containing capabilities of the
support structures. The isolators have a larger mechanical loading than the cables. Foundation of towers have the
tower as the mechanical loading. By design, the transmission towers are expected to fail before the other elements.
 Fig. 1. Mechanical loading transmission lines civil engineering designed. Failure containment. Image adapted from [12].

This paper considers the fragility of towers as independent as in [13]. This assumption is supported by [14] where
the reliability of the elements of transmission lines were evaluated in the field. A transmission line was equipped
with multiple sensors to evaluate mechanical stresses. The stresses were created using explosive charges on the
cables and towers and the stresses on the nearby towers were evaluated. The evaluation supports the independent
consideration of tower failures.
The failure probability of transmission towers is associated with the 3-second averaged wind speed commonly
referred to as wind gust [6], [15]–[17]. The failure rate model for towers was developed with the data from ten years
of named storms that reached the Texas coast, and is presented in [16]. Fig. 2 describes the failure rate of existing
towers that have a wind load requirement of 105 mph in accordance with the National Electrical Safety Code
(NESC) before 2008. Fig. 3 illustrates the failure rate of the towers after 2008. After 2008, NESC incorporated the
American Society of Civil Engineers 7-98 standards – “Minimum Design Loads for Buildings and Other Structures”
[18], which has “extreme” wind and ice loading requirements. Thus, the towers designed with the updated standards
have wind load requirement of 130 mph. The equation describing the failure rate (λ) in relation to the wind gust
(w) from Fig. 2 and 3 is,

λ(w) = αeδw . (1)

where, α and δ are the curve fitting parameters. These parameters for the curve fitting are presented in Table I.
Reference [6] converted the failure rate from [16] to failure probability utilizing the Poisson random sampling.
The authors sample a Poisson distribution with the failure rate to obtain the probability of damage. To avoid using
a random sample, the probability of the Poisson distribution is calculated,

λk e−λ
P (k) = . (2)
k!
 Fig. 2. Failure rate of towers designed with a wind loading of 105 mph (i.e., existing towers). Image extracted from [16].

Fig. 3. Failure rate of towers designed with a wind loading of 130 mph (i.e., hardened tower). Image extracted from [16].

TABLE I
FAILURE RATE CURVE FITTING COEFFICIENTS FOR EXISTING AND HARDENED TOWERS .

Tower wind loading requirement (mph) α δ

−7
105 2 × 10 0.0834
130 2 × 10−8 0.0834

The failures with one or more occurrences are of interest. Thus, the Poisson probabilities of interest are all P (k)
except for k = 0. Since k is non-negative integer numbers, the Poisson probabilities of interest are calculated,
 P (k ≥ 1) = 1 − P (k = 0) = 1 − e−λ . (3)
The tower failure rate from [16] are converted to failure probability using (3). Fig. 4 presents the fragility curve
for both existing & hardened transmission towers for wind gust speed in mph & m/s. Note that the towers designed
with a wind loading requirement of 105 mph and are more likely to fail at lower wind speeds than the tower designed
for the wind loading of 130 mph. This is expected given that the updated considerations of transmission tower
standards require the structures to be able to withstand stronger winds. Which in turn increases the survivability of
the towers to the physical stresses of hurricane winds.

Fig. 4. The fragility curve for the towers designed with 105 mph, and 130 mph from the Fig. 2 and Fig. 3 respective tower failure rate.

III. H URRICANE W IND M ODELING

The wind speed intensity can be estimated utilizing historical hurricane data. Historical hurricane wind information
is available from the National Hurricane Center [19]. Fig. 5 presents the Hurricane Maria path and wind intensity
at each location. Geographic information system (GIS) is used to map the wind speed data to tower locations.
 Fig. 5. GIS map of hurricane Maria’s trajectory and its wind speeds.

To map the wind speed data to tower locations is necessary to interpolate values from the available wind data
to the extent of the study area. The wind swaths represent the footprint in which wind speeds of a certain bin are
expected to occur, for example 64 knots, 50 knots, and 30 knots. Fig. 6 illustrates the wind swaths for Hurricane
Maria over Puerto Rico. It is assumed that wind speed intensity within each swath is at least equal to or lesser than
the wind speed intensity for that swath. Second and third order equations, depending on the number of coincident
swaths - were used to develop linear interpolation between center point measurement of wind speed, and the outer
edge of each wind swath. An interpolation is used to generate 1 X 1 meter resolution wind data.
 Fig. 6. Hurricane Maria wind swaths over Puerto Rico.

The interpolated wind speed is averaged over one minute. To convert the wind speed from one minute average
to three second average wind speed (i.e., wind gust) is not trivial but can be approximated. This conversion of
wind speed to wind gust was done using the area terrain roughness, which is dependent on the surrounding terrain
difference and the elevation around every tower. Using GIS information of the terrain around every tower, the wind
gust factor was computed as presented in [16].
IV. C ALIBRATING F RAGILITY C URVES
The calibration of transmission tower fragility curves could be performed with different approaches. For example,
using the tower design standards, and historical damage reports. The objective of calibrating fragility curves is to
better represent the system under study. As presented in Section I, the fragility curves are the starting point for
multiple studies of HILP events. The accurate representation of the element’s ability to survive HILP events will
impact resilience planning and thus in the end affect the actual system status and response capabilities. Independent
of the type of study: 1) Monte-Carlo simulation to evaluate the power system resilience, 2) Cost-benefit evaluation
of upgrades, 3) Planning for an HILP event, etc. all studies use fragility curves [20].
Reference [16] utilized the empirical method with detailed damage reports from multiple events over the course
the ten years. Their study had complete knowledge of the tower population, towers that failed, and their wind design
rating also considering the changes that naturally happen through time. Such abundance of data is not common
given the challenging and/or dire circumstances the HILP events impose on the region. Given such challenges, the
approach presented in this paper does not create a completely new fragility curve, but it calibrates an existing one
by using one detailed damage report.
If the fragility curve is known, the intensity of weather parameter for every asset is associated with their respective
failure probability. Historical data from events like hurricane Maria, which had numerous element failures, can be
utilized to calibrate fragility curves. The event is required to have cause failures of the elements being calibrated
in order to be utilized. With the combination of the intensity of weather parameter for every asset and with the
associated damage report the calibration of one fragility curve parameter Ψ can be performed. The three-step
interactive procedure describes the calibration of fragility curves, is presented below:
 1) With the intensity of weather parameter for every asset compute the individual failure probability.
2) Having the failure probability for every asset compute the number of expected asset failure.
3) Compare the number of asset expected to fail with the known number of asset that actually failed.
a) If the comparison in Step 3 is above the tolerance, the fragility curve coefficient Ψ is updated and
returned to Step 1. The coefficient Ψ is increased or decreased depending on the previous curve having
overestimated or underestimated the number of asset failures.
b) If the comparison in Step 3 is below the tolerance, the asset fragility curve coefficient Ψ has already
been calibrated and the interactive procedure has concluded.
The interactive procedure for calibrating one parameter of the fragility curves utilizing damage report results in a
calibrated empirical fragility curve.
V. R ESULTS
The calibration of transmission tower fragility curve is performed with damage report for hurricane Maria. The
intensity of the event resulted in numerous transmission towers that failed (not considering smaller damages that
did not lead to tower failure). From the damage report it was evident that the tower construction material (i.e.,
wood, steel, aluminium, and concrete) had a significant impact on probability of failure of the tower. To that end,
the transmission tower fragility curves are calibrated for all different tower material.
Considering that the δ coefficient for tower failure rate (1) is a constant. The α coefficient can be calibrated as in
Section IV. In order to calibrate the transmission tower wind gust fragility curve, the maximum wind gust for every
tower must be known. The maximum wind gust is extracted from a historical event as described in Section III.
The calibration procedure from Section IV is performed for every tower material. The resulting calibrated tower
fragility curve are presented in Fig. 7.
The failure rate coefficients for existing towers is presented in Table II. The wind loading for towers with different
materials is estimated based on the known design standards used for constructing the towers. The failure rate of
the existing and hardened towers at their respective loading requirement is 0.00127 and 0.00102, respectively. The
estimated wind loading of towers is an integer value chosen with the range of possible wind values that result in
a failure rate between 0.00127 and 0.00102. Computing the wind loading with the knowledge of the failure rate is
performed using (4).

ln αλ


w= . (4)
δ
TABLE II
C ALIBRATED FAILURE RATE COEFFICIENTS FOR TRANSMISSION TOWERS OF DIFFERENT MATERIALS .

Tower estimated wind

Material α δ
loading requirement (mph)
Wood 80 1.39 × 10−6 0.0834
−7
Steel 104 1.83 × 10 0.0834
Aluminium 105 1.73 × 10−7 0.0834
Concrete 112 9.44 × 10−8 0.0834

VI. C ONCLUSION
The presented methodology for calibration of fragility curves using limited real-world data is capable of tuning
generic fragility curves to a specific region with their own characteristics. The proposed approach requires the
modeling of the intensity of the weather parameter in the fragility curve for an HILP event, and the knowledge of
the number of elements that failed from the damage report. The presented method is utilized for the calibration of
transmission towers based on their material, however, the calibration can be based on other characteristics of failure
(e.g., structure design standards) depending on the available data. The calibration was performed considering the
main island of Puerto Rico and tested on the towers that failed in each area.
Fig. 7. Calibrated fragility curves for the towers made with different materials. Wind loading for towers with different materials is calculated
based on the known design standards used for constructing the towers.

R EFERENCES
[1] K. Porter, “A beginner’s guide to fragility, vulnerability, and risk,” Encyclopedia of earthquake engineering, vol. 2015, pp. 235–260,
2015.
[2] A. Veeramany, S. D. Unwin, G. A. Coles, J. E. Dagle, D. W. Millard, J. Yao, C. S. Glantz, and S. N. Gourisetti, “Framework for
modeling high-impact, low-frequency power grid events to support risk-informed decisions,” International Journal of Disaster Risk
Reduction, vol. 18, pp. 125–137, 2016.
[3] V. Chalishazar, T. Brekken, D. Johnson, K. Yu, J. Newell, K. Chin, R. Weik, E. Dierickx, M. Craven, M. Sauter, A. Olennikov,
J. Galaway, and A. Radil, “Connecting risk and resilience for a power system using the portland hills fault case study,” Processes,
vol. 8, no. 10: 1200, 2020.
[4] V. Chalishazar, B. Johnson, E. Cotilla-Sanchez, and T. K. A. Brekken, “Augmenting the traditional bus-branch model for seismic
resilience analysis,” in 2018 IEEE Energy Conversion Congress and Exposition (ECCE), 2018, pp. 1133–1137.
[5] B. Johnson, V. Chalishazar, E. Cotilla-Sanchez, and T. K. A. Brekken, “A Monte Carlo methodology for earthquake impact analysis
on the electrical grid,” Electric Power Systems Research, vol. 184, p. 106332, Jul. 2020.
[6] E. B. Watson and A. H. Etemadi, “Modeling Electrical Grid Resilience Under Hurricane Wind Conditions With Increased Solar and
Wind Power Generation,” IEEE Transactions on Power Systems, vol. 35, no. 2, pp. 929–937, 2020.
 [7] L. Ma, P. Bocchini, and V. Christou, “Fragility models of electrical conductors in power transmission networks subjected to hurricanes,”
Structural Safety, vol. 82, no. September 2019, p. 101890, 2020.
[8] M. Ouyang and L. Dueñas-Osorio, “Multi-dimensional hurricane resilience assessment of electric power systems,” Structural Safety,
vol. 48, pp. 15–24, 2014.
[9] D. A. Reed, C. J. Friedland, S. Wang, and C. C. Massarra, “Multi-hazard system-level logit fragility functions,” Engineering Structures,
vol. 122, pp. 14–23, 2016.
[10] M. T. Schultz, B. P. Gouldby, J. D. Simm, and J. L. Wibowo, “Beyond the Factor of Safety : Developing Fragility Curves to Characterize
System Reliability-US Army Corps of Engineers,” no. July, p. 51, 2010.
[11] C. J. Wong and M. D. Miller, Guidelines for electrical transmission line structural loading. American Society of Civil Engineers,
2009.
[12] J. Xue, F. Mohammadi, X. Li, M. Sahraei-Ardakani, G. Ou, and Z. Pu, “Impact of transmission tower-line interaction to the bulk power
system during hurricane,” Reliability Engineering and System Safety, vol. 203, no. June, 2020.
[13] M. Panteli, C. Pickering, S. Wilkinson, R. Dawson, and P. Mancarella, “Power System Resilience to Extreme Weather: Fragility
Modeling, Probabilistic Impact Assessment, and Adaptation Measures,” IEEE Transactions on Power Systems, vol. 32, no. 5, pp.
3747–3757, 2017.
[14] P. Vincent, C. Huet, M. Charbonneau, P. Guilbault, M. Lapointe, and D. Banville, “Testing and Numerical Simulation of Overhead
Transmission Line Dynamics under Component Failure Conditions,” Conférence Internationale des Grands Réseaux Élect, pp. 1–8,
2004.
[15] P. J. Vickery and P. F. Skerlj, “Hurricane Gust Factors Revisited,” Journal of Structural Engineering, vol. 131, no. 5, pp. 825–832,
2005.
[16] Quanta Technology, “Cost-Benefit Analysis of the Deployment of Utility Infrastructure Upgrades and Storm Hardening Programs,”
Satellite Communications, vol. 3021, pp. 1–108, 2009.
[17] R. Cwece, D. Bernard, Y. Krien, F. Leone, T. Candela, M. Péroche, E. Biabiany, G. Arnaud, A. Belmadani, P. Palany, and N. Zahibo,
“A 30 m scale modeling of extreme gusts during Hurricane Irma (2017) landfall on very small mountainous islands in the Lesser
Antilles,” Natural Hazards and Earth System Sciences, vol. 21, no. 1, pp. 129–145, 2021.
[18] ASCE, “Minimum design loads for buildings and other structures.” American Society of Civil Engineers, 1998.
[19] “Historical Hurricane Tracks.” [Online]. Available: https://coast.noaa.gov/hurricanes/#map=4/32/-80 [Accessed 17 Nov. 2021].
[20] V. H. Chalishazar, “Evaluating the Seismic Risk and Resilience of an Electrical Power System,” Thesis, 2019.