W912EK24R0033 CFD Report.pdf

CENWP-ENC-HD

CFD Appendix

CFD Modeling for Lock 25 Tainter Valve Floatation Analysis

April 2024

Contents Background

Model Overview

Grid Development

Boundary Conditions

Model Validation

Turbulence Model Sensitivity

Hydraulic Validation

Force Validation

CFD Model Results

Friction Estimates and Floatation Analysis

Conclusions and Recommendations

Figures Figure 1 - Lock and Dam 25 Aerial Image Figure 2 – Mesh Refinement Areas Figure 3 - CFD Mesh Cross Section and Valve Isometric View Figure 4 - CFD Boundary Conditions Figure 5 - Turbulence Sensitivity Force Graphs Figure 6 - Turbulence Sensitivity Moment Graphs Figure 7 - Validation Lock WSE from Pressure Sensors Figure 8 - ADCP and Pressure Sensor Installation Figure 9 - ADCP Setup Information Figure 10 - CFD Validation Point Locations Figure 11 - Pressure Validation Comparison Figure 12 - Velocity Validation V4 Figure 13 - Comparison of Flow Patterns for Different Gate Openings Figure 14 - Velocity Validation V5 Figure 15 - Velocity Validation V6 Figure 16 - CFD Velocity Data for V1-V3 Figure 17 - Lock Chamber WSE and Tension Meter Data for Force Validation Figure 18 - Initial Force Validation Results Graph Figure 19 - Force Validation Results Graph with Added Tension Figure 20 - Force Validation Results Final Figure 21 - Model Coordinate System and Valve CG Figure 22 - Time Variability of Forces Figure 23 - CFD TV Forces - X Direction Figure 24 - CFD TV Forces - Y Direction

Figure 25 - CFD TV Forces - Z Direction Figure 26 - CFD TV Moments - X Direction Figure 27 - CFD TV Moments - Y Direction Figure 28 - CFD TV Moments - Z Direction

Tables Table 1 - CFD Run Table Table 2 - Turbulence Sensitivity Force Output Table 3 - Turbulence Sensitivity Moment Output Table 4 - Validation CFD Run Parameters Table 5 - Force Validation Gage Data Compilation Table 6 - Force Validation CFD Boundary Conditions Table 7 - Force Validation CFD Initial Force Output Table 8 - Force Validation CFD Initial Moment Output Table 9 - Initial Force Validation Results Table 10 - CFD Valve Forces Summary Table 11 - CFD Valve Moments Summary Table 12 - Moment Summary for Current Valve Table 13 - Moment Summary for Old Valve

Background Lock and Dam 25 is located in Calhoun County, Illinois, and Lincoln County, Missouri, at approximately Mile 241.4 on the Upper Mississippi River above the mouth of the Ohio River near Winfield, Missouri. It is owned and operated by USACE St. Louis District and utilizes 14 Tainter gates and 3 roller gates to regulate water levels to provide navigable waters in the vicinity of the Dam. Ships and barges currently rely on the Project’s 600 ft long lock chamber to allow navigation past the Dam, but a new 1,200 ft navigation lock is currently being designed for the Project to increase the capacity of ship and barge traffic, while the existing 600 ft lock will be used at an auxiliary lock for recreation traffic. The new 1,200 ft lock is being designed to utilize the existing riverward filling and emptying culvert of the 600 ft lock, including the existing filling and emptying valves. The existing 600 ft lock will only be operated using the landward conduit and valves once the new lock is constructed.

As part of the design process for the new lock, a Semi-Qualitative Risk Assessment (SQRA) was performed to evaluate the risks associated with using the current culvert valve system for the new lock chamber, and an issue-specific risk assessment was held to analyze the risk posed by the potential of the valve to "float". One of the main findings from this assessment focused on a current issue that was identified at the Project: floating of the Tainter Valves under ice flushing operations. Operations staff noted that the current Tainter Valves, which are wire rope-operated, traditionally-oriented (skin plate facing upstream) valves, have occasionally not been able to be closed during ice flushing operations which occur with relatively high head differentials. Thought to be because of hydrodynamic loading on the valve under small gate openings and with the wire ropes unable to provide any force in compression to force the valves closed, the valves would remain near 18 inches open. A smaller head differential or equalization of the lock chamber with the forebay water surface elevation (WSE) was required to allow the valves to close under their own weight. The main concern for this operation from a risk perspective is lack of control of the valves through the entire operation cycle, which could cause harm to lock infrastructure, vessels, or human injury or life loss.

St. Louis District reached out to USACE Portland District, who has recently been utilizing Computational Fluid Dynamics (CFD) modeling to get a better understanding of hydrodynamic forces on their reverse Tainter Valves at both the John Day and The Dalles Dam Projects, to perform a similar analysis and estimate the forces causing the Lock 25 valves to float. This analysis included the CFD modeling, which provides the hydrodynamic forces on the valve under given operations as well as estimates for side seal friction and trunnion friction, which would contribute to the valve’s lack of ability to close under their own weight. A CFD modeling plan was created to emulate ice flushing operations and output force and moment data for the upstream landward filling Tainter valve, which was analyzed along with the valve weight and estimated frictional forces to show the propensity for the valve to float under high head differentials. An estimated weight that could be added to the valve to prevent them from floating was also provided, along with potential other mitigation measures. The valve that was modeled in CFD was also instrumented in the prototype to give pressure and flow velocity data as well as wire rope tension and spool-out, which were used to validate the CFD model hydraulic solution and the friction force estimates.

Model Overview The CFD modeling software selected for this study was Star-CCM+ v. 2210 by Siemens. The software solves the Reynolds-Averaged Navier-Stokes (RANS) equations to predict steady-state and transient flow fields in the model domain. The software uses a body-fitted computational mesh, which was deemed important for this study as the geometry for the valves are complex and require the mesh to be able to resolve this complexity without requiring an inordinate cell density around the valve. Star-CCM+ also can calculate forces and moments due to pressure and shear, which is a requirement for this analysis.

Turbulence is an important parameter in resolving the flow field in the model domain. For this study, the k-epsilon turbulence model was originally used as it is a good compromise between robustness and accuracy for a large range of flow conditions. Understanding that turbulence is time-averaged in CFD models, and that the complex flow fields around the valve will contribute to the hydrodynamic forcing on the valves in the prototype, the intent of this modeling is to get an approximation on the hydrodynamic forcing on the valve and recommend a weight that could be added to the valve to prevent floating. To check for the validity of using a Detached Eddy Simulation (DES) turbulence model, which uses a Large Eddy Simulation (LES) subgrid scale model in areas of detached flow regions and a k-epsilon approach in the remaining domain, both k-ε and DES models were evaluated. Model outputs include water surface elevations, pressures, velocities, forces, moments, and streamlines, as appropriate, to help assess the hydraulic performance of the valve.

Combined with the computing power of the High-Performance Computers (HPCs) as part of the Department of Defense (DoD)-funded High Performance Computing Modernization Program (HPCMP), Star-CCM+ provides a robust and informative tool to investigate a variety of flow conditions.

An aerial image of the 600 ft lock at Lock and Dam 25 is shown in Figure 1. The CFD model incorporates a portion of the forebay, the landward flow culvert intake, Control Valve (CV) #2, the landward chamber’s filling and emptying ports, and the lock chamber itself. Shapes for Miter Gates (MG) 1-4 were included as boundaries for the model, but the gates themselves were not an operational part of the model.

Figure 1 - Lock and Dam 25 Aerial Image

Grid Development The meshes for the CFD model runs were created in Star-CCM+ version 2210.0001 (17.06.008). The development of the model grid parameters used for CFD model runs was an iterative process that involved testing and adjusting grid development strategies, along with incorporating meshing techniques that were used for similar navigation lock models for John Day and The Dalles Dams.

The valve CAD geometry provided for the modeling included all the bolts, nuts, and other components that would be required to manufacture the valve. Star-CCM+ requires geometry that is closed, manifold, and non-intersecting. Due to intersection of the various surfaces in the CAD model, as well as small non-contiguous areas within the valve geometry, it was not feasible to try and fix the number of errors that were present to use it directly in the CFD model. Star-CCM+ includes a surface wrapping feature that creates a new surface mesh around the outside of a CAD geometry with given parameters, which was utilized to convert the CAD geometry into a workable CFD surface mesh. Overall shapes for the bolts, nuts, and other thin surfaces were maintained to provide a more accurate estimate of the hydrodynamic forces acting on the valve.

For reference, to compare with the meshing information below, the valves have a radius of 17.70 ft, a height of 13 ft, and a width of 12.5 ft.

A final computational grid for the model domain was developed using a Trimmer cell (rectangular prism-shaped) mesh with the following parameters:

• Base grid: 10 ft cells, two prism layers, grid aligned with the navigation lock water surface elevation (WSE)

• Refinement of total domain: 1 ft anisometric cell size in the z-direction

• Refinement around the Tainter valve vault: 1 ft cell base size

• Curve Refinement around the Valve: target size of 0.1 ft, with a minimum of 0.05 ft

• Refinement around the upper and lower valve seals: 0.025 ft to 0.1 ft cells

• Refinement valve curves and edges: 0.1 ft, with a minimum of 0.05 ft cells

Refinement areas for the vault, valve seals, and valve curves are shown in Figure 2 below. The refinement around the top and bottom seal of the valve were rotated with the valve geometry and re-meshed, depending on the required valve opening for each CFD run. A cross section through the center of the valve vault showing a representative volume mesh, along with an isometric view of the valve mesh, is shown in Figure 3.

Figure 2 – Mesh Refinement Areas

Figure 3 - CFD Mesh Cross Section and Valve Isometric View

Each model had nearly 9.3 million cells, but the exact number varied slightly depending on the valve opening.

Boundary Conditions This CFD effort did not include any fluid/structure interactions that would allow the valves to move in response to the hydraulic forcing, or movement of the valves causing changes to the flow. Instead, the valves were rotated to their desired positions within the model and then removed from the flow domain, which rendered them as stationary objects in the model. Based on a one-valve filling rating curve at the prototype, forebay and lock chamber WSEs were selected that corresponded to the openings of the valve, and each model run was run to a quasi-steady state condition where inflow into the model was near equivalent to outflow. This approximates a “snapshot in time” of the lock filling operation, where the valve is at a set opening and is experiencing a set head differential. Boundary conditions in the model were developed to match the prototype conditions.

The boundary conditions for the Lock 25 model include:

• A stagnation inlet boundary with hydrostatic pressure set at the required forebay WSEs for each run. This was established at the upstream boundary of the model, upstream of the intake.

• Two pressure outlet boundaries, set on the lock chamber-facing sides of both the upstream and downstream miter gates, set to a hydrostatic distribution based on the anticipated lock chamber WSEs for each run.

• Pressure outlet boundaries set at the top of the valve vault, the upstream and downstream bulkhead slots, and the upstream bulkhead slot of the landward emptying valve.

• Symmetry boundaries for the top of the model in both the forebay and lock chamber areas.

Since Star-CCM+ is a two fluid model, simulating both air and water moving through the domain, all inlets and outlets need to be defined for both fluids. This also requires that any area where air can become trapped needs an air boundary to allow the atmospheric pressure to be maintained. Since all inlet and outlet boundaries allowed the exchange of air and water, the top of the model was not required to pass air and hence was set to a symmetry boundary to effectively create a slip boundary with no friction. These main boundary conditions are labeled in Figure 4.

Figure 4 - CFD Boundary Conditions

All other region boundaries were set as default wall boundaries that simulate the effects of friction, including the intakes, conduit and outlet ports, the Tainter valve, and the lock walls and floor. Flow through the model was self-established based on the valve opening and head differential between the forebay and lock elevation, the initial condition for the model was set to the forebay elevation, and the lock chamber was allowed to drain down to its intended elevation until a quasi-steady state condition was reached.

Four different runs were outlined to try and replicate floating valve conditions, based on information from Project operators and ice flushing operations. Floating conditions were observed at the Project when head differentials were high between the forebay and lock chamber, and valve openings were small. The maximum lift for the lock is 15.5 ft, but it was reported that the valve floats anywhere from

13 ft to 15 ft of differential, which is within the normal range of operations. To try and force the floating condition, the forebay was set to an elevation of 433.6 ft NAVD88, and a 15 ft differential was used. This corresponded to a lock chamber WSE of 418.6 ft, and the Tainter valve openings were varied over a small range from 0.5 ft to 1.5 ft. A run table including openings and valve rotations is presented in Table 1.

Table 1 - CFD Run Table for Valve Floatation

Different settings were utilized to validate the CFD to the prototype-collected data, which are outlined in the Model Validation section below.

Model Validation Due to timing of the instrumentation of the Tainter valve in the prototype, the validation of the CFD model and frictional force calculations were done after the main CFD modeling was completed. This was done with the understanding of the risk of possible rework if changes would have been needed to make the CFD model better align with the validation data.

Two separate validation efforts were conducted – one validation for the hydrodynamic solutions coming from the CFD model, and another validation for the force and moment information along with the estimated frictional forces. Pressure and velocity sensors were installed first on the valve that aided in the hydraulic validation, whereas tension meters were installed on the wire lifting ropes later that enabled the force and friction estimate validation. A sensitivity analysis on the turbulence model used in the CFD model was also conducted.

Turbulence Model Sensitivity K-epsilon and DES turbulence models were evaluated for the force validations runs, to try and understand the effect of each model on the results and determine which one is applicable to this study.

Two runs were selected at small valve openings, two in the middle of the opening cycle, and two at the upper end for comparison. The K-ε turbulence model is one of the most common turbulence models used and is reasonably accurate for a wide range of applications. The DES model emulates a basic LES subgrid scale model in detached flow regions, which could be applicable to the flow separation regions around a Tainter valve, but uses a K-ε turbulence model in the remainder of the domain. It was unclear if the scale of the flow separation around the valve could be resolved by a DES turbulence model, as it usually requires well defined and large-scale vortex shedding in a massively separated flow field, but the comparison helped to understand the effect of the turbulence models on the force and moment results.

Force outputs from both turbulence model runs (for X and Z directions) are shown in Table 2, and moment outputs (for moments about the Y axis) are shown in Table 3.

Table 2 - Turbulence Sensitivity Force Output

Table 3 - Turbulence Sensitivity Moment Output

Based on the results, the estimated forces on the Tainter valve using both a K-ε and LES models were similar for small and large valve openings but varied more around the ~50% opening. This was reflected in the moments about the Y axis, but the small differences in the force information reflected as larger changes in the moment estimates. This is shown graphically for the force information in Figure 5 and for the moment data in Figure 6.

Figure 5 - Turbulence Sensitivity Force Graphs

Figure 6 - Turbulence Sensitivity Moment Graphs

The force information is relatively similar using both turbulence models for the smaller gate openings, which are the openings that are of concern for the floatation of the Tainter valves. Even though the force magnitudes are close, the DES turbulence model runs showed a higher magnitude negative moment about the Y axis, which is representative of the valve trying to float. Because of the higher magnitude and to be conservative with the estimates, it was decided that the main four runs for estimating floatation would use a DES turbulence model.

Hydraulic Validation The hydraulic validation utilized prototype measurement data to check against CFD model output to verify if the hydraulic solution is accurate. Two different types of gages were installed by the ERDC- CRREL team on the landward-side filling valve of the lock, which matches the valve being modeled in CFD. The main validation is based on a 1 valve filling operation performed on 4October2023, from 7:53am to 8:07am. The forebay elevation at the Lock 25 project during the validation operation was

434.09 ft, and the starting lock WSE was 420.91 ft, which provided a differential of 13.18 ft. This data was provided, along with the valve openings over time for the operation, which made it a good validation data set.

The first gages installed were pressure sensors at three separate elevations – on the invert of the lock valve chamber, 6 ft above the floor of the chamber, and 12 feet above the floor of the chamber. For redundancy, two sensors were located at each elevation. The vertical line of pressure sensors was installed in the ladder well within the valve chamber, which shields them from the high velocity of the flow moving through the system. These data were collected on a frequency of 1 minute and were correlated to the lock chamber WSEs by converting the pressure to hydraulic head in feet. A curve was then fit to the data, and used to approximate boundary conditions for the model, as the valve opened much faster than data was collected. A graph showing the WSE over their duration of the lock filling and emptying cycle is shown in Figure 7. Overall, the three levels of pressure sensors agreed with each other, with a slight variance between the top and bottom sensors. One thing to note is the pressure drop at the beginning of the filling cycle before time 5 minutes that is below the starting tailwater elevation; this is thought to be due to local drawdown in the valve chamber as water started moving under the opening gate. As the WSEs from the pressure sensor were used to set boundary conditions in the CFD model, a linear approximation was used for the WSE when the pressure sensor graph dropped below the tailwater elevation.

Figure 7 - Validation Lock WSE from Pressure Sensors

The other sensor that was installed was an Acoustic Doppler Current Profiles (ADCP), which uses sound waves to measure velocities in two separate beams across an area and averages them to get an approximate velocity in a straight line perpendicular to the instrument. The ADCP was installed at the same elevation as the middle pressure sensor, 6 ft off the floor of the valve vault, directly downstream of the ladder well and pointing across the valve vault. Pictures of the ADCP installation, as well as the pressure sensors, are shown in Figure 8.

Figure 8 - ADCP and Pressure Sensor Installation

The ADCP uses two beams, at a 20-degree angle to each other, to collect data and provides velocity measurements in both the X and Y directions in 20 cm bins across the valve vault. The recorded data are the average of the velocity vectors from both beams, and due to the final bin extending through the wall, only 17 bins were recorded from the gage. Setup information for the ADCP is shown in Figure 9.

Figure 9 - ADCP Setup Information

Due to inherent limits to the ADCP, data were only output on five-minute intervals (the gage could only collect and process the large amounts of data every five minutes). This became a limiting factor for validating the CFD model through the one valve filling cycle, as the valve reportedly opens within a 1 minute and 45 second duration. Because of this, velocity data to validate the CFD model to the prototype only aligned with the time when the valve was fully open but before the WSEs equalized between the forebay and lock chamber. To account for this limitation, six CFD runs were planned to utilize both velocity and pressure data for validation. Those runs are outlined in Table 4 below. Note that the times next to each run represent the prototype time that the sensor data were collected.

Table 4 - Validation CFD Run Parameters

Only pressure data was available to validate Runs 1-3, and both pressure data and velocity data were utilized for Runs 4-6. Probes were inserted into the CFD model for each level of pressure sensor, as well as the center of each ADCP bin across the valve vault, and data were extracted over the duration of each run, which allowed a comparison of minimum, average, and maximum data for each point. This is important for trying to compare to prototype data taken at a specific point in time, as the flow profile is believed to be time-variable, even if the flow under the gate is constant, because of the turbulent nature of flow downstream of the valve. A diagram of the different probes in the CFD model is shown in Figure

10. Note that there are 18 probes in the CFD model replicating the ADCP bins because the probes were located in the centroid of the bins. There is also velocity data in the vertical (Z) direction as an output from the CFD model, which isn’t available from the prototype data but can still be used for insight into the flow profile downstream of the valve.

Figure 10 - CFD Validation Point Locations

The spread between the minimum and maximum pressures from the CFD model were minimal, so only average pressure values were compared to the prototype values. Graphs comparing the CFD output to the prototype data are shown in Figure 11. Note that location z1 is the lowest sensor, and Z3 is the sensor highest in the water column.

Figure 11 - Pressure Validation Comparison

Overall, the pressure data validated well. The pressures are consistent between the CFD and prototype for V1, V5, and V6, but there is some variability with the other 3 runs. For V2, it looks like the prototype is recording higher than the CFD model, which is exaggerated with V3. There was some concern with understanding where the valve is in its opening cycle vs. what is reported by the Human-Machine Interface (HMI) indicators, and this could be a result of not having the correct valve opening set for the given WSE differential. Also, using the pressure data from the prototype to prescribe water surfaces in the lock chamber doesn’t account for the head loss through the conduits and ports, which would correlate to a different pressure value in the CFD model. Even with those two factors, the values were close and it indicates that there may be some inconsistencies with valve position when looking at the first velocity comparison.

The velocity information was compiled from the CFD model for V1-V6, and adjusted to be in the same coordinate system as the prototype data. Only runs V4-V6 are used for comparison to the prototype data, as the ADCP time data frequency didn’t fall within the duration of the valve opening (Figure 7 shows ADCP data collection times). All runs were with the valve fully open, and only the lock chamber WSE was varied per the pressure data. Velocity data was graphed as vectors utilizing the X and Y magnitudes on top of an image of the CFD model with the measurement points active, overlayed with the prototype data from the ADCP. After comparing V4-V6, some general flow characteristics will be described from the velocity data for V1-V3. Output data for run V4 are presented in Figure 12. Similar to the pressure data, the range between minimum and maximum velocities is small for the CFD results, so the averages were used to compare to the prototype.

Figure 12 - Velocity Validation V4

The magnitude of some of the vectors is similar, but the direction is very different between the CFD model and prototype data. The valve is supposedly fully open in the prototype at this time, but it’s showing recirculation across the width of the valve vault and as well as some vectors pointing upstream.

For a fully open valve, it would be assumed that most of the flow would be uniform and generally moving downstream, as the gate lip is out of the flow path and not causing any flow separation. Figure 13 shows a vertical cross section with velocity magnitudes (utilizing different scales) for runs V3 and V4, with the valve still controlling flow in V3 while the valve is fully open in V4. There is a large shear zone between the fast-moving water underneath the valve and the low velocity water above at the plane of flow separation, which would induce recirculation similar to what is shown in the prototype data for V4.

Because of these, it is assumed that the valve wasn’t fully open when it was reported as open during the validation operation. This assumption will be investigated using the information from the tension meters.

Figure 13 - Comparison of Flow Patterns for Different Gate Openings

The velocity comparison for V5 is much closer than V4 and is shown in Figure 14. Velocity magnitudes are generally in agreement for the main points in the middle of the valve vault, except for a fairly large magnitude in the prototype at point 2. There is some small difference in flow directions between points, but all points are directed downstream. Due to the flow direction, it is assumed that the valve is fully open in the prototype for this comparison and that the CFD model is replicating the prototype well.

Figure 14 - Velocity Validation V5

For the final comparison at V6, the CFD output matches the prototype data well (see Figure 15). This comparison has the lowest magnitude velocities since the head differential between the forebay and lock chamber is the lowest, and all flow is directed downstream. There are some slightly higher velocity magnitudes towards the center of the conduit from the ADCP data, but overall the CFD results match well with the prototype data.

Figure 15 - Velocity Validation V6

CFD velocity information for Runs V1-V3 are presented in Figure 16, including a table of vertical velocity components for each point. Points are numbered from the bottom of the image up, and vertical velocity is labeled as Data_k in ft/sec. The smallest gate opening shows low magnitude velocity components, due to the flow moving under the gate and not having much affect on the water column 6 ft above the floor of the valve vault where the ADCP is measuring data. Velocity data for V2 shows a recirculation in the flow path, where flow is moving upstream in the left portion of the flow path (looking downstream) and moving downstream in the right side. There is also some variability in the vertical component of the flow as well, meaning that there is probably rotational flow passing under the valve causing both vertical and horizontal eddies. V3 shows flow all moving downstream, but a large variability in the magnitude of the velocity possibly due to the interaction of the gate lip with the flow passing underneath it. The vertical velocity distribution for V4 is opposite of V3, where in V3 the outsides of the flow path were directed downward and the interior was up, in V4 the outsides are directed up and the interior is directed down. CFD results show a twisting flow coming into the valve from the intake in the forebay, and this could be a continuation of that twisting flow past the valve.

Figure 16 - CFD Velocity Data for V1-V3 (blue vectors are CFD results, orange dots are measurement locations)

Force Validation The intent of the force validation is to use information from tension meters installed on one of the lifting cables to validate which hydrodynamic forces are being used in the estimate (min, max, avg) and adjust the calculated frictional forces for both side seal friction and trunnion friction. Both sources of friction contribute to the valve not being able to close under its own weight, and they aren’t parameters that are able to be modeled in CFD. The tension meters also have a “spool out” reading, which will be used to validate the location of the valve during its operation in relation to lock chamber WSEs and tension forces in the cables.

A 1-valve fill operation was used for the force validation, and the operation to collect the data was completed on the 12January2024, near 8:20 am CST. The forebay WSE was 434.35 ft NAVD88, and the starting lock chamber WSE was 422.36 ft NAVD88. I took 1 min and 37 seconds to fully open the valve.

It should be noted that the conduit height it only 12.5 ft, but the valve was reported to open to 14.5 ft total. The lock chamber filled 2.38 ft as the valve was opened, and it took another seven minutes to fill the remaining 9.61 ft so that the chamber WSE matched the forebay and the head was equalized.

Before the valve began to operate, the tension meter reported an average force of 3.88 kips, which keeps the lines in tension and keeps the wire ropes wrapped tightly around the hoist drums. Once the valve was lifted to 14.5 ft open and the head equalized, the tension meter reported an average force of

8.22 kips.

Pressure data was pulled from the pressure sensors to correlate with the valve opening time period, to set up the CFD model validation. The pressure data was converted to static head and used to estimate the water level in the lock chamber over the duration of the valve operation. Figure 17 shows the water surface elevation and tension data for the operation.

Figure 17 - Lock Chamber WSE and Tension Meter Data for Force Validation

Eleven valve openings were selected to analyze in the CFD model, to compare the force estimates to the measured tension data. Multiple smaller openings were selected to capture the tension changes as the valve is pulled off-seal, and then 2 ft opening differentials were used for the mid-operating range. The upper range captured the point when the valve seal was above the conduit ceiling at 12.5 ft, and the reported full open of 14.5 ft. A table showing the various openings, corresponding time (in both Central

Standard time (CST) and Alaska Standard Time (AKST), as the tension data and pressure data were recorded in different time zones), and force and lock chamber WSE is shown in Table 5.

Table 5 - Force Validation Gage Data Compilation

A table of the CFD boundary conditions is also shown in Table 6.

Table 6 - Force Validation CFD Boundary Conditions

Models were run out near 500 seconds of model time, but the models appeared to converge after 100 seconds. Forces and moments, with a reference coordinate system shown in Figure 21, were compiled for the last 100 seconds of run time, and statistically analyzed for the minimums, averages, and maximums over time. The results, focusing on forces in the X and Z directions, along with moments about the Y axis, were used to estimate side seal friction and trunnion hub friction based on standard design calculations. Forces and Moment data are shown in Table 7 and Table 8.

Table 7 - Force Validation CFD Initial Force Output

Table 8 - Force Validation CFD Initial Moment Output

Friction estimates were converted to moments and summed about the trunnion hub axis along with the hydrodynamic moment reported by the CFD model, and the moment from the weight of the valve. The resulting moment was analogous to the moment caused from the tension in both wire ropes, which divided by the radius of the valve (and then by two to get tension in one rope) would be comparable force value for the tension meter data. The main variables that effect the friction forces are the valve preset for the side seal friction and the trunnion friction coefficient for the trunnion hub friction. The original intent was to use these coefficients to “tune” the friction estimates to match the tension meter data. This adjustment would then be used for the floating valve friction estimates to provide more confidence in the magnitude of force estimates preventing the valve from closing. Initial results for calculated moments and wire rope tension estimate are shown in Table 9, and a graph of the comparison to the gage data is shown in Figure 18.

Table 9 - Initial Force Validation Results

Figure 18 - Initial Force Validation Results Graph

As can be seen in the graph, the tension meter data corresponds well with the calculations for smaller valve openings (0.5-1.5 ft). As the valve is opened further, the initial estimates for tension appear to be underpredicted, with the calculated tension falling within the 20% error band only for the 7 ft open case.

For the large openings, the original calculations seemed to underapproximate the tension by ~3kips in a single wire rope for openings greater than 5 ft. This is a large enough difference that it prompted more investigation into the calculations, and ultimately the gage data.

Per the original validation plan, the coefficients for seal preset and trunnion friction coefficient were modulated to see if it would bring the calculations closer to the gage data. It was found that these values have little effect on the overall tension in the wire ropes, as the high hydrodynamic moment-controlled results for the initial opening of the valve, and the reduction of the hydrodynamic force and increase of the effect of the weight controls for the larger valve openings. As the results were close for the lower gate openings, it appeared that it was the effect of the valve weight that were differing from the gage data. The gage data was investigated further to see if there was explanation for the differences in the data itself.

The portion of the tension gage data that could be analyzed without the influence of friction forces or hydrodynamic forces was the reported tension when the valve was fully open, and the head was equalized between the forebay and lock chamber. With the valve fully open and no influence from other forces, the wire ropes and the trunnion hub should be accounting for the entire wet weight of the valve (valve weight minus buoyancy). In this condition, summing the moments around the trunnion hub would include the weight of the valve applied at the horizontal distance from the trunnion hub to the center of gravity (CG), which would be offset by the tension in the wire ropes multiplied by the radius of the valve. At a 14.5 ft opening, the horizontal distance from the trunnion hub to the CG is 11.58 ft, and the radius of the valve is 17.7 ft. With an estimated valve wet weight of 16.41 kips, this would result in a total wire rope tension of 12.3k, or 6.15 kips per wire rope. The tension gage data recorded an average value of 8.22 kips under this condition, which is an additional ~2.1 kips of tension in the wire rope than anticipated.

There are multiple reasons that the gage data could be showing a different value than anticipated: the valve is heavier than believed, the CG of the valve is different than what was calculated, the wire ropes could be distributing the force of the valve unequally, or the gage itself may not be fully calibrated.

Without more analysis and investigation, it’s unknown which of these variables are causing the differences, but trying to normalize the difference can help validate the calculations to the gage data.

To try and normalize the discrepancies between the calculations and the gage data, 2.1 kips was added to the calculated tension force, results shown in Figure 19. This brought the calculated tension closer to the gage data, but it appeared that calculations were then over-estimating tension for the smaller gate openings.

Figure 19 - Force Validation Results Graph with Added Tension

All of the initial force validation runs were completed using a K-ε turbulence model, but the DES results from the sensitivity test showed that higher floatation was predicted when using a DES turbulence model. To try and account for the differences in the smaller gate openings between the calculations and the gage data, the DES model results were used for the smaller gate openings (0.5 ft to 3 ft), and it correlated better when the minimum Y hydrodynamic moments were used for openings 0.5 ft to 1.5 ft, and the maximum Y hydrodynamic moments were used for openings 3 ft to 14.5 ft. A final comparison graph is shown in Figure 20.

Figure 20 - Force Validation Results Final

The initial effort to validate the force data to the tension gage information was meant to change friction constants to align the calculated forces with data from the prototype. As the gage data was analyzed, it became apparent that the large magnitude of the hydrodynamic moments in the smaller valve openings controlled, the distribution of the weight of the valve controlled for larger openings, and the friction estimates had minimal effect on the overall tension in the wire ropes for this case.

More effort could be scoped to investigate why the gage data didn’t show anticipated tension when the valve was full open with equalized head; whether it was due to the gage itself, the understanding of the valve, or the distribution of forces between the two wire ropes. Additional effort could also be expended to understand the effect of the turbulence models at various valve openings, and if it was more accurate to vary which model was used based on the opening and boundary conditions. By utilizing the available data and making some assumptions, the calculated tension in the wire ropes was able to mimic the trend of the gage data within a 20% error. Also, the force validation helped to reinforce the use of a DES turbulence model for smaller valve openings, which is the range of openings that is important to the floatation estimates.

CFD Model Results Results from the CFD model included time-varied force and moment data collected on all components of the Tainter valve, along with subsequent flow information and pressure images used for qualitative evaluation of the valve. Force data are summed for every mesh cell face on the valve, in each direction (X, Y, and Z). The coordinate system for both the force and moment information is located with the Y axis through the center of the trunnions, the X axis facing upstream, and the Z axis opposing the direction of gravity. A CG was also computed, based on the valve geometry in the model, for use in the floatation analysis. A representation of the model coordinate system, along with the CG information, is shown in Figure 21.

Figure 21 - Model Coordinate System and Valve CG

The forces are inherently calculated in Star-CCM+, based on both pressure and fluid shear. Pressure plots of the valve show pressure concentrations along the surface of the valve, and the cumulative force data give magnitude to the overall hydrodynamic force acting on the valve, but the tabular outputs don’t describe any force concentration locations on the valve since they are summed over the entire valve. The moment data, which are defined based on the force data, are the most useful in understanding force concentrations and was used for the floatation analysis to understand the hydrodynamic forcing on the valve. Star-CCM+ does not inherently allow for calculation of the moments but allows for user-input code (field functions) to describe mathematical functions within the CFD model. A field function was developed by Siemens to calculate the moment based on the forces and uses the cross-product method for moment calculations. The moments taken about each axis are then summed for each cell to get a cumulative moment on the valve for all directions of motion.

Even though all models were run to quasi-steady state conditions, there are still time-varied hydrodynamic conditions in the modeling domain due to the tortuosity of the approach in the conduit and fluid interactions with the valve geometry itself. Figure 22 shows forces on the valve for Run 1, with a 0.5 ft valve opening. After 1,000 model seconds of time, the forces seem to be in an equilibrium condition compared to the initial run time where the model is trying to move from the initial conditions into a physics-based solution. All model output data was statistically analyzed over time for minimum, average, and maximum values after the models hit their quasi-steady state conditions. By comparing the spread of data over time, the variability of the forces and moments can be described. As determined in the turbulence sensitivity and the force validation efforts, a DES turbulence model was used for these CFD model runs.

Figure 22 - Time Variability of Forces

Forces from the CFD model are presented in Table 10 and moments are summarized in Table 11.

Table 10 - CFD Valve Forces Summary

Table 11 - CFD Valve Moments Summary

To better understand the relationships of forces and moments, as well as the variability in the data, graphs for each direction and variable of interest were compiled. For reference, large negative forces in the X direction would indicate high pressure on the skin plate of the valve, whereas high positive forces in the Z direction would indicate propensity of the valve to try and open due to hydrodynamic forcing. Y direction forces can give some insight into if the valve is trying to rack itself laterally, but tended to be very small compared to forces in the X and Z directions. As the moments were taken around the trunnion hub, halfway between each trunnion, the main emphasis for this analysis is the Y moment, which indicates the valve rotating around the trunnion hub with a large negative force corresponding to the valve attempting to float. Both X and Z moments can give some insight into the torsion on the valve, but both were orders of magnitude smaller than Y moment output. Force charts are shown in Figure 23 through Figure 25, and moment charts are shown in Figure 26 through Figure 28.

The forces on the valve through the four runs shows a propensity for the hydrodynamic forces to try and keep it open. There is a large magnitude force in the X direction (with flow in the conduit) pushing on the skin plate, as well as a large magnitude force in the Z direction. Both forces decrease as the valve is opened, but combined they try to force the valve to rotate open around the trunnion hub. This is also evident in the Y direction moment, which shows a net opening force on all four valve openings. Forces in the Y direction, as well as moments around the X and Z axis are variable throughout the runs, but they are an order of magnitude lower than the forces in the X and Z direction and moments around the Y axis. These “racking” forces and moments may be due to the twisting of flow as it approached the valve from the forebay intake.

Overall, the forces from the CFD model are showing that the hydrodynamic forces are trying to keep the valve floating with high head and low gate openings, but frictional forces and the effect of the gate weight needs to be analyzed to get a full understanding on how the valve would react under these conditions. These estimates are provided in the next section.

Figure 23 - CFD TV Forces - X Direction

Figure 24 - CFD TV Forces - Y Direction

Figure 25 - CFD TV Forces - Z Direction

Figure 26 - CFD TV Moments - X Direction

Figure 27 - CFD TV Moments - Y Direction

Figure 28 - CFD TV Moments - Z Direction

Friction Estimates and Floatation Analysis To get an accurate understanding of how the valve would react under the hydrodynamic forces from the CFD model, a free body diagram was used to sum all the moments about the trunnion hub. This includes the moments due to the hydrodynamic forces, the valve weight at the center of gravity, the side seal friction, and the trunnion hub friction. Note that the friction forces were taken to resist the valve from closing, and therefore oppose the weight of the valve. Also, the wire rope resistance was not added because the moments weren’t summed to a static condition, and if the valve was floating, the wire ropes wouldn’t be in tension. The following equation was used to check for overall moments on the valve (Equation 1):

Equation 1 - Sum of Moments About Trunnion Hub

Calculations for the side seal friction were modified from examples provided by CENWP-ENC-DM and reference EM 1110-2-2107, and calculations for the trunnion hub friction were also modified from examples provided by CENWP-ENC-DM and reference EM 1110-2-2702. Friction force estimates and methodology were checked by the Lock 25 PDT, and the calculations with comments will be provided as an attachment to this report.

For the first draft of the calculations, a seal preset of 0.125 in and a trunnion friction coefficient of 0.15 was used as they were believed to be realistic values for the prototype valve. To be conservative, a seal preset of 0.25 in and a trunnion friction coefficient of 0.3 was used for the friction calculations. Also, previously the estimated buoyant weight of the valve was used in the moment calculations, but the full dry weight of the valve should be used in the calculation; the hydrodynamic force from the CFD model inherently includes buoyancy, so using the buoyant weight of the valve would incorrectly overestimate the buoyant pressure on the valve.

Moments for each force acting on the valve, as well as the resulting moments, are presented in Table 12 below.

Table 12 - Moment Summary for Current Valve

With the full summation of moments on the valve, the 0.5 ft valve opening with the 15 ft head differential showed a propensity for the valve to float with a moment of 15,371 lbf-ft. The other runs showed a net closing of the valve, which would result in loading the wire ropes in tension and would allow the valve to close under its own weight.

It is believed that the original valves utilized at Lock 25, which were heavier, also experienced floatation under similar circumstances. As a check on the calculations, they were also put together using the weight and dimensions of the old valve. The old valve weighed 23,518 lbs, had a radius of 17.07 ft, and a horizontal distance to it’s CG of 9.99 ft. Calculations are shown in Table 13.

Table 13 - Moment Summary for Old Valve

The calculations for the old valve did not show a resulting negative Y moment for any of the openings, which means the valve should have closed under its own weight. Even though none of the openings showed floatation, the resulting moment for the 0.5 ft opening was relatively small, and increased friction beyond the calculated values or dynamic flow changes as the valve was closed could potentially cause the valve to float even with the additional weight.

Conclusions and Recommendations To better understand the floating valve condition with the Tainter valves at the Lock 25 Project, a CFD model was created to estimate the hydrodynamic forces on the valve under large heads and small gate openings. These forces were compared to the other forces acting on the valve, including the weight of the valve and friction from the trunnion hub and side seals, to numerically replicate the prototype valve and understand the amount of force necessary to offset the floating condition seen at the project during ice flushing operations. An extensive validation effort was undertaken, including a CFD model turbulence sensitivity, comparing CFD output to prototype-collected data from pressure sensors and ADCP measurements, and also comparing to tension meter data from the wire ropes operating the valves to tune the frictional force estimates.

Findings showed that there were significant hydrodynamic forces, both acting on the valve against the skin plate as well as vertically on the valve, which contribute to the valves inability to fully close during ice flushing operations. Based on the friction estimates and contributions from the weight of the valve, only the CFD-modeled condition of a 0.5 ft valve opening with a head differential of 15 ft caused a high enough hydrodynamic force that it offset the resisting forces and would cause the valve to float. With this condition, the sum of moments around the trunnion hub was 6,308 lbf-ft in the direction of maintaining the valve open. The resisting forces used in the analysis are considered conservative. The side seal friction calculations utilized a side seal preset value of 0.25; it is unlikely that the seal preset value was this large based upon inspections of the valve.