Investigation of Thunderstorms on an Instrumented Building

1
Investigation of Thunderstorms on an Instrumented
Building
Wind Speeds and Wind Induced Pressures
Franklin T. Lombardo, Ph.D. Candidate Douglas A.
Smith Texas Tech University April 25, 2008
2
Project Summary/Problem

• The engineering properties of wind, no matter
  the source, are homogenous ? ASCE 7
• Thunderstorm flow in its own regard has been
  shown to be different than that of its
  non-thunderstorm counterpart (wind speed
  increases, direction changes, etc)
• Differences in wind induced pressures however,
  have not been fully investigated
• can be done with WERFL
• If differences can be determined this project
  has broad, wide ranging impacts
• Classification of thunderstorm winds in
  structural code
• Parameterization
• Difficult to characterize wind engineering
  parameters (GF, TI, etc due to nonstationarity)

• Less damage to low-rise buildings
• Thunderstorms dominate wind climates in most
  temperate areas of the United States and the
  world
• Less damage to buildings, over/under design

3
Introduction

• Impinging jet profile
• Maximum mean wind speed likely between 25 70 m

• Homogenous properties of wind
• fundamental assumption of wind engineering
  practice
• currently not adequately suited to handle
  thunderstorm winds (Orwig and Schroeder, 2007,
  Choi, 2002)
• wind pressures expected to vary with wind speed,
  turbulence integral scale, turbulence intensity
• Thunderstorm flow itself has differences from
  non-thunderstorm
• rapid increase(s) in wind speed
• rapid change(s) in wind direction
• convectively driven
• vertical velocity
• nonstationary (largely)
• BL profile

July 6, 2004 (WERFL)
Martner, 1997
4
Introduction (Cont.)

• Since properties are expected to be different,
  it is important to analyze as a separate
  phenomenon
• Thunderstorm winds generate large economic
  losses (Crandell et al., 2000)
• Control design for temperate climates (Holmes,
  2001)
• Will wind induced pressures exhibit different
  qualities?
• HOW DO WE DO IT???
• Time Domain
• Pressure Coefficients (Average, Conditional)
• Correlation Analysis
• Wind Vector Analysis (Vertical AOA)
• Time/Frequency Domain
• Wavelet Analysis (PSD, Scalogram)

5
Thunderstorm Building Effects

• No documented full-scale thunderstorm effects in
  literature
• Many wind tunnel simulations of
  downburst/microburst
• Simulation of thunderstorm effects
• Nicholls et al. (1993) ? LES
• large negative Cps at leading edge of roof
• Sarkar et al. (2006)
• Cps were larger than TBL
• uplift loads exceeded design specifications
• Letchford (2002)
• Large negative Cps in wind tunnel
• Simulation of tornado (Selvam and Millett, 2003,
  2005)
• vertical wind component highly concentrated when
  striking the building
• recirculation of vorticity produces highly
  localized suction on roof

6
Thunderstorm Building Effects (Cont.)

• Simulation of tornadoes ???
• strong vertical velocities
• horizontal, vertical vorticities
• also present in thunderstorm
• Vertical Velocity
• large negative pressures as is increased ()
  (Richards and Hoxey, 2004)
• peak suction pressures found with large ()
  excursions in component for conical vortices (Wu
  et al. 2001)

May be weak, but if vorticity is stretched
(gustnado)
7
Thunderstorm Building Effects (Cont.)

• Secondly, colder and denser air may cause
  increased load on structures (Chay and Letchford,
  2002)
• Along with denser air, comes rapid increase in
  wind speed, change in wind direction
• rapid increase in loading
• wind direction diverges from mean azimuth by
  large angles is typically where peak pressures
  occur (Letchford and Marwood, 1997)
• Flow may also have higher lateral correlation
  (Holmes, 2008, Sarkar and Sengupta, 2008)
• Four important characteristics
• vertical velocity and associated vorticity
• higher density/pressure air
• increase(s) in wind speed (profile difference as
  well)
• change(s) in wind direction

Seminole, TX 2007 (Marc Balderas)
8
Thunderstorm Building Effects (Cont.)
3-4 Increase
9
WERFL Building

• Instrumented Building
• 200 pressure taps (walls, roof)
• Wind speeds (tower, sonic on building)
• To calculate pressure or Cp includes calculation
  of mean dynamic pressure
• Differential pressures are calculated at each
  tap and divided by q to generate Cps. Due to
  nonstationarity of thunderstorms ? erroneous
  value of Cp as it is largely dependent on wind
  speed. However p can be retrieved
• This pressure information can then be used for
  further analysis

10
Detailed Comparisons

• Recorded 10 thunderstorm runs from Spring 2003
  - Present
• In order to compare with non-thunderstorm runs,
  it is ideal to
• use similar angle of attack runs ( 5 deg)
• must use similar terrain profiles (Cp varies
  greatly)
• qualified runs have varying building position
• Run 2480 ? July 7, 2004
• Mean AOA 351 deg (STAT) ? suburban terrain
  (150 m)
• Maximum wind speeds 70 mph (31 m/s)
• 21 non-thunderstorm qualified runs with these
  characteristics
• No sonic available ? all NT runs had operational
  sonic
• 13 anemometer used
• 150 ft (46 m) from building
• Traditional (correlation), Non-Traditional
  (wavelet) to accommodate additional
  non-stationary events

11
Run Comparisons
Thunderstorm(11908)
Non - Thunderstorm(11908)
Wind Speed (m/s)
Wind Speed (m/s)
Run 2480
11.8 m/s
13.1 m/s
AOA (Deg)
AOA (Deg)
9-13
347 deg
350.8 deg
Pressure (Pa)
Pressure (Pa)
44.4 Pa
83.1 Pa
47.88 Pa 1 psf
12
Thunderstorm Run (Cp, Rvp)
thunderstorm
15 (4.6 m)

1
5

• Lessen dependence on mean wind speed in
  calculation of Cp(t), stationary records
• Thunderstorm flow has higher running mean Cp
• May mean higher correlation with incident
  pressures

13
Thunderstorm Run (P Correlation)
15 (4.6 m)

• Thunderstorm run appears higher correlation
  (avg, indiv) for most taps for this AOA (windward
  wall)
• Lateral correlation may be higher for
  thunderstorm flow ? shown in thunderstorm
  full-scale wind speeds

14
Thunderstorm Run (P Correlation)

• Similar normalized profile
• Relationship between lateral TI and normalized
  pressure at first tap (degree of separation)

15
Power Spectral Density Estimation

• Method employed in Gurley and Kareem (1997)
  using wavelet estimation (non-stationary events)
• sum of squares of wavelet coefficients at
  different frequency bands
• wavelets extract mean value of non-stationary
  history and turbulence at different frequency
  levels, or bands remain
• selected windward and roof tap

• Energy seems shifted to slightly higher
  frequencies

16
Other Thunderstorm Runs

• Run 2933 ? May 20, 2003 (sonic operational)
• 11 comparison runs (similar AOA)
• Computation of vertical angle of attack
• Weaker event ? peak winds 50 mph (15 m/s)
• Use of conditional sampling
• Lessen effect of wind speed
• Scalogram (Run 0003/0004 ? August 24, 2007)
• Identifies time/frequency components
• Ramp up/directional shift case

17
Vertical Angle of Attack

• Noted in Richards and Hoxey (2004), Wu (2001)

• 11 NT runs
• May 2003

• Although fluctuations in w component are
  significantly larger, the subsequent changes in
  other components equalizes vertical AOA (other
  runs as well)

18
Conditional Sampling

• 15-min run times using mean dynamic pressure
  likely not adequately suited for investigating
  small-scale differences in pressure (where
  comparison runs may not be available)
• Method employed by Wu (2000)
• calculate instantaneous pressures p(t)
• Then recalculate pressure coefficients using
  instantaneous velocity vector V(t) (or mean
  vector) for some designated period of time (1s,
  60s etc) from 30 sonic (no sonic for some
  events)
• Helps to isolate data to determine if there are
  any significant effects from factors other than
  wind speed occurring at relatively small scales
  (3-s) ? discretize AOA
• Windward Wall, Roof
• Assume air density is constant

19
Conditional Sampling

• Limited AOA variation (lt 10 degrees over 3-s
  span)
• Thunderstorm coefficients are within bounds of
  their NT counterparts
• Weaker event, longer fetch ? may be differences

20
Scalogram

• August 24, 2007 event (shown earlier) ? ramp
  up/directional shift
• Identifies transient behavior in the
  time/frequency domain
• Shows frequency bands (scales) that accompany
  these events
• increase in wind speed/pressure ? higher
  frequency scales

21
Thunderstorm Pressure Summary

• For similar mean angle of attack and wind speed
• Some events display different BL profile
• Possibly higher correlated with thunderstorm
  wind speeds (Sarkar and Sengupta, 2008)
• Mean Cps in stationary thunderstorm records
  higher than all comparison records
• Higher correlation among pressure taps (windward
  wall)
• Power spectral density estimates (Gurley and
  Kareem, 1997) show
• Different packets of energy than NT pressures
• Vertical velocities
• Run shown and other runs display similar
  characteristics in vertical AOA although w
  component increases
• Vorticity introduction may be different at
  thunderstorm leading edge or in internal
  structure (flow visualization)

22
Thunderstorm Pressure Summary

• Conditional sampling will eliminate effects of
  wind speed to enable focus on other factors
• flow correlation, etc (Quasi-Steady Theory)
• Strong frontal boundaries (higher correlation)
• May have significant differences from event to
  event
• Strength of thunderstorm winds
• Important to identify transient/non-stationary
  behavior of thunderstorms (wavelets, other
  techniques)
• Fundamental question
• Does thunderstorm data (wind pressures, etc)
  display readily identifiable characteristics
  apart from non-thunderstorm data?
• Future Study
• Continue to collect thunderstorm pressure/wind
  data throughout spring/summer

23
Questions/Comments
24
(No Transcript)