Overview

This page describes experiments in a water tank, part of a set of controlled chamber experiments used to investigate pollutant dispersion in large spaces.

The water tank is a 30:1 scale model of a full-scale test chamber. It was designed to:

• yield denser data sets, both in time and in space, than the full-scale facility;
• help with the design of experiments in the full-scale facility;
• perform experiments that would be too time-consuming in the full-scale facility; and
• provide data for validating Computational Fluid Dynamics simulations of pollutant dispersion in a large space.

Concentration fluctuations during one run of the water tank.

This page describes the following aspects of the water tank experiments:

• The water tank design, including its physical layout and visualization techniques.
• Transient results, showing how pollutant fills in to the room under a steady flow field.
• Fully-developed results, showing how random fluctuations in the flow field affect concentrations even after the pollutant has filled in to the room.
• The effect of obstructions on the pollutant flow.
• Scaling theory, explaining how the experiments relate to full-scale tests.
• References for further information.

Return to top

Water Tank Physical Layout

The water tank consists of a 30:1 scale model of a full-scale test facility, housed in a 20-gallon container. Clean water flows into the model room, through the supply inlets in one corner, at approximately one tank volume per minute (about 28 lpm). The incoming water mixes with the water and dye in the tank, then flows out the top of the tank. The upper surface of the tank is free, with no lid or ceiling.

Schematic diagram of the water tank.

Both the flow inlet and outlet positions, and the flow rate, were chosen to mimic the conditions in the full-scale facility. While it is not possible to match the full-scale facility exactly, scaling theory explains how the experiments relate to full-scale tests. The interior arrangement of the tank varies between experiments, in order to examine the effect of different room configurations on the room flow. Experiments include runs with the space unobstructed, with model-scale tables, and with model-scale obstructions representing people. The human-like obstructions are not heated, and therefore do not simulate the effect of the thermal plumes generated by real people. The model itself occupies a 20-gallon container, divided by a partition. One side of the partition houses the scale model, while the other side provides a reservoir from which to remove water exiting at the exhaust vent. Once it flows out, water does not recirculate back into the model.

Return to top

Visualizing the Water Tank Flow

To visualize the flow of pollutant through the water tank, a fluorescent tracer dye (sodium fluorescein with a concentration of 10 mg/l) flows from a foam ball situated just above the tank floor. A peristaltic pump maintains a constant dye flow rate of 1 cc/s. Due to scaling effects in the water model, this corresponds to a release rate of about 1.8 l/s in the full-scale test facility.

The dye mixes with the water, showing its flow patterns as the water rises through the model and exits through the overflow weir.

A video camera, mounted directly above the tank, records the light emitted when the dye is excited by an argon laser. To observe the dye concentration in the breathing plane, the laser light is focused into a 1 cm thick sheet through a plane 6 cm above the tank floor. The intense blue-green laser light causes the dye to fluoresce, re-emitting light at a different wavelength. The dye fluoresces in direct proportion to the dye concentration, and to the local intensity of the light sheet.

The images from the camera are digitized at up to five frames per second for video, or one frame per second for higher-resolution still images. These still images have a maximum resolution of about 320 by 240 pixels; since the image extends slightly beyond the edges of the tank, this means each pixel represents an area of about 0.1cm by 0.1cm.

In addition to experimental runs, the camera records images under background conditions (i.e., with no dye present), and for several different well-mixed dye concentrations. These measurements allow image processing in order to account for variations of light intensity in the plane, and to calibrate the system for different dye concentrations.

The image processing also smooths the recorded frames to reduce the effect of camera noise. The smoothing process averages the pixel values in successive, non-overlapping squares, five pixels on a side.

Return to top

Water Tank Transient Results

To show the evolution of the pollutant distribution over time, the images below show the concentration in the breathing plane, starting from the initial release of tracer dye in a fully-developed flow field. The tank contained no obstructions.

Since each run of the experiment represents only one realization of a highly variable and stochastic process, it requires multiple runs to adequately represent the spread of dye over time. To obtain a well-defined profile for the growth of the pollutant plume, the figures above represent pixel-by-pixel averages over 19 separate runs of the experiment. That is, each image represents the average of 19 concentration images, each from a different run, but taken at the same elapsed time after the start of dye injection.

The flow field for every run was established at least three minutes before injecting the tracer dye. At a flow rate of one tank volume per minute, this ensured at least three volume changes before starting the experiment.

Return to top

Water Tank Fully-Developed Results

The transient results above show how the dye fills in on the breathing plane of the water tank. After running a continuous dye release for some time, the dye concentration reaches a fully-developed state, in which the time-averaged concentrations are stable.

However, even after the concentration distribution becomes fully-developed, the stochastic nature of the flow results in large changes in the instantaneous concentration over time. The figure below compares two frames from a single run of the water tank. The differences between the two snapshots result from random variation in the flow, rather than from a meaningful change in the dye distribution.

When averaging a small number of frames, the concentration profile depends on the particular frames used for the averaging. Thus the pair averaged across five frames shows obvious differences between groups 'A' and 'B'.

As a larger number of frames enters the averages for the two groups, their images converge to a consistent average concentration profile. For a large enough sample, the average values should be identical. The figures suggest that using 200 frames in each average provides an accurate picture of the fully-established average tracer dye concentration in the measurement plane.

Return to top

Effect of Obstructions in the Water Tank

To study the effect of obstructions in the full-scale facility, we compared flow visualizations for water model configurations corresponding to an open room, a room with tables, and a room containing both tables and people. The model representations of the people were not heated, and so do not capture the effect of the thermal plumes generated by real people.

As shown in the figures below, the presence of obstructions increases the overall concentration in the measurement plane, and shifts the areas of peak concentration. Furthermore, obstructions tend to make the tracer dye spread out more before it enters the measurement plane.

These pictures indicate that obstructions of the sort typically found in rooms can have a significant impact on the average concentration in the measurement plane, even in the absence of thermal plumes.

Return to top

Scaling Theory

Scaling down the physical dimensions of the full-scale test facility, and changing the working fluid from air to water, obviously affects the flows of interest. To retain similarity with the full-scale test facility requires adjusting the time scale used to interpret results observed in the water tank. The appropriate scaling transformations depend on the factors affecting turbulence in both the full-scale room and the water tank model.

The experiments described here investigate mechanical ventilation only, with no buoyancy effects involved. Restricting the flows to neutrally buoyant plumes, with isothermal fluid and surfaces, allows us to characterize turbulence in terms of the Reynolds number only (the Reynolds number gives the ratio between inertial and viscous forces acting on the fluid).

Making the Reynolds numbers for flows in the water tank model equal to those in the full-scale room would ensure compatibility of the scale model results. Ideally, the scale model achieves equal Reynolds numbers for all flows of interest (here, the flow at the inlets as well as the general flow within the room). In practice, it is not always possible to achieve an exact match. Therefore we use turbulence theory to estimate the differences between the model and full-scale results.

As run, the model has a Reynolds number approximately 80% that of the full-scale facility. This shifts the time scale to approximately 15:1. That is, one second of time in the water model is equivalent to 15 seconds in the full-scale chamber. While this limits the size of the smallest eddies that the water model can predict, those smallest eddies still fall below the resolution of the camera.

Return to top

References

For more information on the topics on this page, please see the following references. Sources denoted with an LBNL number may be found among the publications of the Airflow and Pollutant Transport Group, or on the Indoor Environment Department publications page.

• Turbulence theory. A.A. Townsend, F.R.S. The Structure of Turbulent Shear Flow, Second Edition. Cambridge University Press, Cambridge G.B., 1976.
   
• Water tank experiments. Thatcher et al. Pollutant Dispersion in a Large Indoor Space: Part 1 - Scaled experiments using a water-filled model with occupants and furniture. LBNL-50248.

Return to top