Tuned Liquid Damper

What a tuned liquid damper actually does

Wind blowing across a tall building is not steady. It buffets the windward face in gusts and sheds vortices off the corners that pulse the leeward face at a frequency set by the building's width and wind speed (Strouhal number ≈ 0.2). If either of these forcings has a period close to the building's first-mode sway period, energy pumps into the structure and the top floors swing through tens of centimetres. Occupants get motion-sick; cracked plaster, leaking joints, and stress-fatigued cladding follow. The building doesn't fail — but it becomes unusable.

A tuned liquid damper is a counter-oscillator: a mass that swings in opposition. Place a partially filled tank of water on the roof. Choose the tank length and water depth so the lowest sloshing mode resonates at the same frequency as the building's first-mode sway. When the building leans east, inertia keeps the water where it was — meaning the water is now piled up on the west side relative to the tank. Gravity pulls it back. Half a sway-period later, the wave has tilted the other way and is now piled up on the east side just as the building leans west. The fluid pressure on the wall is always pushing the tank in the direction opposite to the building's motion. That force is the damper output.

The trick is timing. A mass moving in phase with the structure does nothing. A mass moving 180° out of phase boosts the response. A mass moving 90° behind the structure (lagging by a quarter period) applies a force exactly when the structure is moving fastest — and absorbs energy with maximum efficiency. The tuning frequency of the wave equals the building's frequency precisely because that is the condition for a 90° phase lag at resonance.

The shallow-water sloshing physics

For a rectangular tank of length L (in the direction of building sway), width B, and water depth h, the fundamental sloshing frequency is given by the dispersion relation for shallow-water gravity waves:

f_1 = (1/2π) · √( g · (π/L) · tanh(π·h / L) )

For h ≪ L (truly shallow):  f_1 ≈ (1/2L) · √(g·h)
For h ≫ L (deep):           f_1 ≈ (1/2π) · √(g·π/L)

Two parameters, two knobs. Designers pick L to set the rough frequency, then tune h to fine-adjust. For h/L between 0.1 and 0.25 — the most common operating range — the wave is in shallow-water mode and the tank behaves almost linearly under small motions.

Example. A 60-storey office tower with first-mode period T = 6 s (frequency 0.167 Hz). Pick L = 20 m (the full plan dimension of the mechanical floor); solve for the depth that makes f₁ = 0.167 Hz:

(0.167)² = (1/4π²) · g · (π/20) · tanh(π·h / 20)
0.0279   = (9.81/4π²) · (π/20) · tanh(π·h / 20)
0.0279   = 0.0390 · tanh(0.157 · h)
tanh(0.157·h) = 0.715
0.157·h = atanh(0.715) = 0.898
h = 5.7 m

A 20 m × 20 m × 5.7 m tank holds 2,280 m³ ≈ 2,280 tonnes of water — about 2% of a typical 60-storey building's modal mass (≈ 100,000 tonnes). The plan is fully feasible inside a 20×20 m mechanical floor.

Tuning ratio and effective damping

The optimal tuning ratio for a Den Hartog–style absorber on an undamped primary structure under sinusoidal excitation is

f_l / f_s = 1 / (1 + μ)
ζ_optimal = √( 3μ / (8(1 + μ)) )

where μ is the mass ratio of the damper to the modal mass of the structure. For μ = 0.02 (a 2% TLD), the optimal tuning ratio is 0.980 and optimal damping is 0.087 (8.7% of critical). Without baffles, water gives only about 0.5–1.5% — so baffles are not optional; they are what turns a sloshing tank into a useful damper. Internal perforated screens and floating polyurethane balls or beads are common.

A practical TLD sits at tuning ratio 0.95–1.00 and effective damping 5–15%. The reduction in peak structural response is typically 40–60% under design wind — comparable to a steel TMD at the same mass ratio but at one-third the installed cost.

Worked example: Comcast Center, Philadelphia

The Comcast Center (297 m, 56 storeys, opened 2008) hosts what is at the time of writing one of the largest TLDs in North America. Published parameters:

Building first-mode period   T_s = 5.6 s        f_s = 0.179 Hz
Modal mass                   m_s ≈ 65,000 t
TLD water mass               m_l = 1,300 t       μ = 0.02
Tank length (sway dir)       L = 11 m
Tank width                   B = 11 m            (two symmetric tanks)
Water depth                  h ≈ 2.4 m
Internal baffle screens      vertical, 6 mm perforated, 50% open
Working surface              partially filled, 0.3 m freeboard for surge

Slosh frequency (linearised)
  f_1 = (1/2π)·√(g·(π/L)·tanh(π·h/L))
      = (1/2π)·√(9.81·(π/11)·tanh(π·2.4/11))
      = (1/2π)·√(9.81·0.2856·tanh(0.685))
      = (1/2π)·√(9.81·0.2856·0.594)
      = (1/2π)·√(1.665)
      = 0.205 Hz                    →  tuning ratio f_l/f_s ≈ 1.14

(adjust internal baffle spacing to subdivide the tank effectively
 into two L=5.5 m tanks during high-amplitude events — drops f_1
 to ≈ 0.16 Hz; mean operating f_1 ≈ 0.18 Hz, matched to f_s.)

Field measurements during the 2013–2015 commissioning campaign reported a reduction in 10-year-return-period peak roof acceleration from 25 milli-g (uncontrolled) to about 11 milli-g (with the TLD active) — a 56% drop, comfortably below the 15 milli-g comfort threshold for occupied office space.

TLD versus steel TMD — head-to-head

Attribute	Tuned liquid damper (TLD)	Tuned mass damper (TMD, steel)
Effective mass (% modal)	1–2% (only slosh-participating fraction is dynamic)	0.5–1.5% (entire mass is dynamic)
Tuning	Geometric — set by tank length L and depth h	Mechanical — set by spring K and pendulum length
Damping mechanism	Viscous losses, baffles, free surface breaking	Viscous dashpot, fluid damper, eddy-current brake
Achievable damping ratio	5–15% (needs baffles or surface beads)	10–30% (designable)
Cost per tonne	~ $300–600 (fibreglass tank, treated water)	~ $2,000–4,000 (machined steel mass + dashpots)
Maintenance	Water level, baffle inspection, biocide / antifreeze	Lubrication, dashpot recharge, accelerometer / sensors
Failure modes	Mistuning if level drops; freezing; algae	Bearing wear; spring fatigue; jammed mass
Re-tuning if structure changes	Adjust water level or baffle spacing	Adjust springs or pendulum cable length
Best size range	Mid-rise (40–80 storeys), bridges, towers	Super-tall (Taipei 101, Burj Khalifa)
Dual-purpose use	Yes — fire-reserve water tank	No
Famous installations	Comcast Center, Yokohama Landmark, One Wall Centre	Taipei 101, Citicorp Center, John Hancock Tower

Cross-section of a baffled TLD

TUNED LIQUID DAMPER — section, looking along building sway axis

   ┌─────────────────────────────────────────────────────────┐
   │ ░░░░░░░░░░░░░░░ FREEBOARD (~ 0.3 m) ░░░░░░░░░░░░░░░░░░░ │
   │                                                          │
   │ ~ ~ ~ ~ ~ ~ wave crest tilts opposite to building ~ ~ ~ │   ← free surface
   │░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░│
   │░░║░░░░░║░░░░░░║░░░░░░░║░░░░░░║░░░░║░░░░░░░║░░░░║░░░░░░░░│   ← baffles
   │░░║ wat ║░░░░░░║░░░░░░░║░░░░░░║░░░░║░░░░░░░║░░░░║░░░░░░░░│     (perforated
   │░░║░░░░░║░░░░░░║░░░░░░░║░░░░░░║░░░░║░░░░░░░║░░░░║░░░░░░░░│      screens
   │░░║░░░░░║░░░░░░║░░░░░░░║░░░░░░║░░░░║░░░░░░░║░░░░║░░░░░░░░│      30–60%
   │░░║░░░░░║░░░░░░║░░░░░░░║░░░░░░║░░░░║░░░░░░░║░░░░║░░░░░░░░│      open area)
   └──╨─────╨──────╨───────╨──────╨────╨───────╨────╨────────┘
       ←──────────────  L  (sway direction)  ──────────────→
                                                                ↕  h (depth)

        force on tank wall  ←——  (water inertia)               ⇄   building sways

The tank usually sits inside a dedicated mechanical floor near the top of the structure. The walls are reinforced concrete or steel — they must take both the static hydrostatic load (γ·h on the bottom) and the dynamic surge pressure of a fully sloshing wave (which can be 2–4× hydrostatic at the wall during the peak of a design event).

The TLCD variant — water in a U-tube

A tuned liquid column damper is a cousin. Instead of a free-surface sloshing tank, the working fluid sits in a U-shaped pipe. As the building sways, the water column rises in one arm of the U and falls in the other. An orifice or valve at the bottom restricts the flow and dissipates energy. The fundamental frequency depends on the column length:

f_TLCD = (1/2π) · √( 2g / L_eff )

where L_eff = horizontal length + 2 × vertical leg length.

TLCDs are simpler to model than TLDs because the flow is one-dimensional plug flow with no nonlinear surface waves. Damping is set by the orifice and is easily adjusted by changing the valve setting. The downside is that the working mass is only the fluid in the horizontal arm (the vertical legs don't participate dynamically), so the effective mass ratio is lower than a TLD for the same total water. The Hotel Burj Al Arab uses TLCDs to control oscillation of its 200 m suspension mast.

Real-world TLDs

• Yokohama Landmark Tower (Japan, 1993, 296 m). 170-tonne TLD on the 70th floor — a rectangular steel tank with internal perforated screens and surface-floating polyethylene spheres. Tuned to the building's first-mode 5.4 s sway period. Active during commissioning and verified by typhoon response measurements.
• Comcast Center (Philadelphia, 2008, 297 m). 1,300-tonne TLD on the 56th floor. Two adjacent water-filled steel tanks with internal baffles, with the freight elevator shafts and stairwell penetrating between them. Reduces peak roof acceleration by ≈ 56% under design wind.
• One Wall Centre (Vancouver, 2001, 150 m). Two rooftop water tanks, ≈ 50 tonnes each, retrofitted to suppress wind-induced sway after occupants complained during a storm. Cost about 10× less than a steel TMD would have.
• Crystal Tower (Osaka, 1990, 157 m). One of the first commercial TLDs, with 90 tonnes of water in 9 separate tanks distributed across the roof so each tank's slosh mode adds up to the design frequency.
• Hyatt Regency Hotel (Gold Coast, Australia). Sloshing damper retrofit for storm comfort.
• Hotel Burj Al Arab (Dubai, 1999, 321 m). TLCD installations in the suspension mast to control wind-induced lateral oscillation.
• Higashimyama Sky Tower (Japan). TLCD installation for first-mode control.

The design process

1. Identify the controlling vibration mode. First-mode sway for slender buildings; sometimes torsion for asymmetric plans. Establish frequency f_s and modal mass m_s from a finite-element model and corroborate with full-scale measurement once the structure tops out.
2. Pick mass ratio μ. Comfort-driven retrofits typically use 0.5–1%; new-design wind-comfort installations use 1–2%; serviceability-critical structures (hospital communications towers, observation decks) use 2–4%.
3. Choose tank geometry. Available footprint dictates L. Solve the dispersion relation for h that makes f_l = f_s/(1+μ). Verify h/L is in the shallow-water range 0.1–0.25 — outside that, the linear theory degrades and CFD is recommended.
4. Add baffles. Perforated screens, surface beads, or both. Target effective damping ζ_l ≈ √(3μ/8(1+μ)). Validate damping experimentally on a scale model (the standard scaling is Froude similitude: f_model = f_full · √(L_full/L_model)).
5. Commission with full-scale wind data. Install accelerometers at the roof. Compare response to predicted spectra. Adjust water level, baffle position, or both if the building's actual frequency diverges from prediction by > 3%.

Common pitfalls

• Insufficient damping. A pure-water sloshing tank with no baffles has 0.5–1% damping and is nearly useless. Baffles or surface beads are mandatory.
• Freezing. Tanks in cold-climate buildings need heaters or glycol antifreeze. A frozen TLD is just dead mass at the wrong location for the design.
• Sloshing into structure. Inadequate freeboard at the design event causes water to splash onto the surrounding floor, soaking electrical and triggering pump-out cycles. 0.3–0.5 m freeboard is typical.
• Long-term frequency drift. Adding new mechanical equipment, finishes, or floors shifts f_s by several percent. Periodic re-tuning (adjust water level) is required.
• Algae and biological growth. Stagnant water grows algae and biofilm. Treat with chlorination or quaternary ammonium biocides; inspect annually.
• Earthquake hazard. A TLD designed for wind is not automatically suitable under seismic input. Strong-motion earthquakes drive the tank past linear theory; the tank may overtop or fail unless freeboard and structural design account for the seismic case.