This article discusses that plant engineers at the James W. Jardine Water Purification Plant in Chicago tweak the treatment process to improve its efficiency and effectiveness. Engineers later studied computer models of flow patterns in the intake area. Using Fluid Flow Analysis software from Algor Inc. of Pittsburgh, the engineers determined a better location for the activated carbon feed point and avoided the expense in time and money of traditional laboratory testing. Department of Water filtration engineer Anthony Wietrzak turned his attention to the mixing of the carbon with the water. To do it, he had to study the water flow in the intake basins. Running several variations on the pump cell model gave Wietrzak the opportunity to experiment with different model constraint techniques and gauge the effectiveness of those techniques in terms of model convergence. Wietrzak has also modeled other areas of the plant that may benefit from modifications.
The James W. Jardine Water Purification Plant in Chicago processes nearly one bill ion gallons of water on an average day. Chicago's Department of Water believes Jardine is the largest water treatment plant in the world. That plant and one other, the South Water Purification Plant, serve nearly five million consumers in the city and 118 outlying suburbs.
With so much water to purify, plant engineers tweak the treatment process to improve its efficiency and effectiveness. For instance, a few years ago, the Jardine plant decided to change the location at which activated carbon is added to the water. Technicians began to add carbon at an earlier stage of the process, in the intake basins, to give it more time to work.
Engineers later studied computer models of flow patterns in the intake area. Using Fluid Flow Analysis software from Algor Inc. of Pittsburgh, the engineers determined a better location for the activated carbon feed point and avoided the expense in time and money of traditional laboratory testing.
The purification process takes about seven hours during which the water passes through a number of steps, including chemical treatments at various stages. Several additives, among them carbon, chlorine, and fluoride, are added to the water in the plant's intake basins. Chlorine, whose job is to kill pathogens, can be added at multiple points throughout the process. Fluoride is an additive to prevent tooth decay.
Polymer is also added in the intake basin to act as a coagulate for alum, which is added later.
From the intake basins, batteries of pumps raise water about 22 feet, or 7 meters, above the lake surface, and gravity carries the stream through its purification. Pumping stations distribute the final product to the city and its suburbs.
The alum, added more than an hour after the water has been pumped into the plant, causes particles of impurities to stick together so they will drop out faster in the settling tanks. This sedimentation phase removes about 90 percent of the particle matter from the water before it passes through filter near the end of the process.
The last additive, polyphosphate, is a material that coats the Surfaces of pipes as it passes through them, and prevents the lead in old plumbing from leaching into the drinking water.
Activated carbon, now among the early chemical treatments, originally entered the channels of water very close to the point where the alum is introduced, well over an hour into the process.
Carbon's purpose is to remove objectionable tastes and odors. The tiny carbon particles are tremendously absorbent, like sponges. The longer they stay in the stream and the more thorough their dispersion, the better the job they do in making the water pleasant and palatable.
In the late 1990s, engineers for the Chicago Department of Water determined that the activated carbon feed should be moved to the intake basins, which take water directly from Lake Michigan. Introducing the carbon earlier in the process would make it more effective by giving it more time to work.
Every spring, plant workers drain and clean each of the two intake basins. Work crews have only a short window of opportunity each year to clean the basins and make any needed modifications. The activated carbon feed line was extended and mounted on the inside wall of each intake basin, over a small ledge-a location that was selected for convenience of installation.
The modification achieved the goal of increasing the time that carbon spent in the water. Next, Department of Water filtration engineer Anthony Wietrzak turned his attention to the mixing of the carbon with the water. To do it, he had to study the water flow in the intake basins.
Working Back from the Pump
To analyze the flow pattern in the intake basin, Wietrzak had to start with the physical characteristics of the system that he knew quantitatively: the capacities of the pumps that pull water from the intake basin into the plant and the physical dimensions of the intake area.
Wietrzak did not want to assume that the flow was uniform coming out of the intake basin. That assumption could result in an unrealistic flow pattern, because velocity could vary as much as 10 percent, depending on the elevation of the water in the basins.
The alternative, modeling both the pump cells and the intake basin, would result in a very large model, which Wietrzak's computer hardware, an Intergraph TDZ2000 workstation, might not be able to process in the time set aside for the project. So he decided to model and analyze one large pump and the intake basin separately.
The results at the inlet boundary of the pump cell model would determine the input for the intake basin model. He would assume that a secondary, smaller pump would have a similar velocity profile.
The pump is a mixed-flow type made by Fairbanks Morse, which is a unit of BF Goodrich based in Beloit, Wis. The pump is rated to handle 320 million gallons a day and entered service in the 1960s.
Wietrzak modeled the volume within a pump cell in Algor's Superdraw In as 6,492 solid brick elements. He applied zero velocity constraints to the surfaces of the pump cell's walls. To the free surface of the fluids he applied a zero shear constraint.
Next, Wietrzak converted the pump capacity of more than 1.1 billion liters, per day to a volumetric flow rate in cubic feet per second. The flow rate was then applied to the pump cell outlet as a velocity boundary constraint. No constraints were placed on the inlet to the pump cell.
An unsteady fluid flow analysis was performed on the pump cell model with the applied velocities ramping up over 50 time steps.
"The largest assumption I made for the pump cell model is that the flow is uniform coming out of the pump," said Wietrzak. "I ran several variations of the pump cell model in which I varied the outlet velocity constraints at the pump discharge, and none made a significant difference in the velocity profile at the inlet to the pump cell. All models yielded higher velocities at the top of the entrance and lower velocities near the bottom.
Since I was concerned only with the velocity profile boundary condition at the outlet of the basin, I am satisfied that this is a sensible assumption."
Running several variations on the pump cell model gave Wietrzak the opportunity to experiment with different model constraint techniques and gauge the effectiveness of those techniques in terms of model convergence. "I know the theory and how to handle problems with a textbook approach. However, there were several constraint techniques that I had to learn to run this model and get realistic results," said Wietrzak, a professional engineer who holds a doctorate in fluid dynamics from Northwestern University. "For example, the free surface boundary condition had to be properly applied to get the solution to converge." He submitted questions to Algor's technical support team.
Armed with the results of the pump cell flow analysis, Wietrzak was ready to tackle the intake basin model. He modeled the volume in the basin in 8,138 solid brick elements. As with the pump cell model, Wietrzak applied zero velocity constraints to the surfaces of the basin walls and zero shear constraints to the free surface of the fluid. The velocity results of the pump cell analysis were then applied to the area where the basin connects to the pump cells. An unsteady fluid flow analysis was performed on the intake basin model with the applied velocities ramping up over 50 time steps
Wietrzak then sliced through the model layer by layer and studied the simulated flow patterns throughout the model. He also used animated analysis replays to see how the solution progressed over time.
He discovered a recirculation pattern next to the wall, very near the feed point.
"The feed point is located above a narrow ledge," Wietrzak observed. "The flow pattern predicted that the carbon would tend to be dragged along the ledge rather than mixing quickly and thoroughly with water throughout the volume of the basin. Our observations of the carbon deposits qualitatively confirmed the fluid flow analysis results."
Wietrzak thus determined that the activated carbon feed pipe should be moved out from the wall between 5 and 10 feet, or roughly 1.5 to 3 meters, to maximize mixing. This modification was implemented in one of the two intake basins in the spring of last year.
The slurry containing the carbon flows into the water from the open end of a pipe. According to Wietrzak, the amount of carbon added can vary from 20 to 80 pounds per million gallons of water, depending on lake conditions and the recommendations of a weekly tasting panel.
Wietrzak has also modeled other areas of the plant that may benefit from modifications. For example, since completing the activated carbon feed point project, he has studied the effectiveness of the air scrubber system on exchanging the air in the chlorine battery room, from which chlorine is supplied for the treatment process. The scrubber releases caustic gas to neutralize airborne chlorine. Wietrzak used fluid analysis to find the dead spots in the airflow