3D Navier-Stokes Module Applications
The first application using the AdH flow solver is for the Marmet Navigation Lock intake. The Marmet Lock and Dam are located on the Kanawha River in West Virginia. The computational model of the intake included 215 m of the upper approach, the upper miter gates and recesses, the intake ports, and filling system culverts. The surface of the tetrahedral mesh (Figures 1 and 2) shows the details of the discretization near the intake ports. As flow approaches the intake, it crosses over an emergency bulkhead sill and enters the six intake ports located on the face of the miter gate sill. The intake splits the flow into two culverts each having three intake ports. The 4.0 m-wide by 4.5 m-high culverts simultaneously circle around the miter gate pintles and drop down to the lock floor elevation through complicated horizontal and vertical curves. The flow accelerates as it enters the intake ports. Further accelerations occur as the port flows combine into a single culvert on each half of the structure. These accelerations and resulting pressure variations are shown in Figure 3 and Figure 4 which are contours plotted along a plane located in the center of the culvert beginning at the port faces. Comparisons with observed laboratory data are shown in Figure 5 in the form of a pressure coefficient, Cp, plot. Here, the reference pressure and velocity, used to compute the pressure coefficient, is zero at upper pool elevation and is the average velocity in the culvert, respectively. The model reproduced the pressure distribution within the structure quite well. The only large discrepancy between the computed and observed values is at piezometer LI5. The laboratory data is suspect at this location, compared to the other observed pressures, due to a faulty pressure tap.
Figure 1. Surface mesh of lock intake.
Figure 2. Surface mesh at the port faces.
Figure 3. Velocity contours on a plane through center of intake and culverts.
Figure 4. Pressure contours on a plane through center of intake and culverts.
Figure 5. Comparison of pressure coefficients from physical and numerical models.
Another intake model is presented to show how model results can highlight design weaknesses. The particular project is the Poe Lock, located on the St. Marys River in Sault Ste. Marie, Michigan. The Poe Lock intakes are 8 port intakes located in each wall just upstream of the lock miter gates. The flow ribbons on Figure 6 are the computational solution at the lock hydrograph’s peak discharge. The ribbons show that the first three ports are essentially useless in that no flow is seen entering these ports. Design refinement could have led to a more efficient intake in which the flow is more uniformly distributed along the face of the manifold. The detail of the computational model provides the necessary insight required to optimize hydraulic design given project constraints.
Figure 6. Flow ribbons illustrating streamlines into Poe Lock intake.
The final application example is the simulation of a complete lock filling operation. The project is Webbers Falls Lock, on the McClellan-Kerr Arkansas River Navigation System about 5 miles (8 km) northwest of Webbers Falls, OK. A typical lock chamber on the Arkansas River is nominally 110 ft (33.5 m) wide by 600 ft (183 m) long. The lock culvert system is a sidewall port design. A model of the Webbers Falls Lock, which has a 29-ft (8.8-m) lift at normal pools, is shown in Figure 7.
Figure 7. CAD model of Webbers Falls Lock filling system.
The model reproduces approximately 500 ft (150 m) of the upstream approach, twin sidewall intake manifolds, the lock chamber, valve wells, sidewall ports and culverts. The culvert system consists of 8-port 7.5-ft (2.29-m) wide by 10-ft (3.05-m) tall intake manifolds on either lock wall from which the flow transitions to 12-ft (3.66 m) by 12-ft (3.66 m) culverts in each wall. Flow is introduced into the lock chamber by 14 ports in each sidewall filling and emptying manifold. Each port is 3.5 ft (1.07 m) tall and has a throat width of 2.54 ft (0.774 m).
The initial finite element mesh has 229,090 nodes and 1,126,845 tetrahedral elements. The mesh resolution is illustrated by the surface mesh shown in Figure 8. Flow variations are resolved during flow simulations using mesh adaption. The mesh adaption consists of refinement and coarsening. The mesh is not allowed to become coarser than the initial mesh. Therefore, the minimum number of nodes and elements that occur during a simulation are those describing the initial mesh.
Figure 8. Computational surface mesh of lock filling system with water surface removed.
The ten minutes of filling operation were simulated using a 5-sec time step. Flow variations within the chamber during filling are shown in Figures 9-11. These pictures show velocity contours at particular times during the simulation. The lock chamber water surface rises from the lower pool (initial conditions) to the upper pool as the lock fills. The jets issuing from the manifolds 1 min after filling started are shown in Figure 12. These velocity contours provide an insightful view of the flow distribution and jet interaction in the culvert manifold and chamber.
Figure 9. Velocity magnitude contours in the lock chamber 1 min after filling started.
Figure 10. Velocity magnitude contours in the lock chamber 4 min after filling started.
Figure 11. Velocity magnitude contours in the lock chamber 8 min after filling started.
Figure 12. Velocity magnitude contours showing flow in each sidewall manifold and lock chamber.