Modeling: Physical Models: Why Hydro Developers Use Them in the Computer Age

Although computational fluid dynamics (CFD) modeling is used extensively in the hydropower industry, some physical phenomena cannot be accurately predicted through CFD. Examples of recent modeling work illustrate the current trend of using physical and CFD modeling in tandem to provide trustworthy results.

Simulating hydraulic behavior using a model is a tried-and-true method for determining how a proposed hydroelectric facility will interact with its environment. Modeling can be the standard design method or may be recommended when other methods are not applicable or do not provide adequate information. But in this era of two- and three-dimensional computer models, why is a large, scaled physical model ever necessary? And what can be gained from such a study?

Computational fluid dynamics (CFD) modeling increasingly is being used in the hydro industry. Despite the growing sophistication of this tool, some physical phenomena still cannot be accurately predicted using CFD. Physical modeling methods provide an additional trust factor because they are mature, have been used to develop many of the hydraulic design elements currently employed, and are palpable and visible to all observers.

Developing trust in the model

It is common perception that CFD modeling is faster and cheaper than physical modeling. In many cases, this is true. In fact, certain types of computer simulations routinely are used in design processes, including turbine runners and draft tubes. However, there are many cases, particularly for situations involving complex geometries, where physical models are more than competitive in schedule and cost.

One primary function of modeling is to test alternatives. In this regard, CFD models do not always enjoy an advantage over physical models. For example, it often is possible to use readily available materials – such as plywood, sheet metal, pipe sections, bricks, and gravel – to make changes while a physical model is operating, sometimes in a matter of hours. For a CFD model, modifications of this nature would require remeshing and rerunning the model, which may take days or weeks. On the other hand, significant modifications that require breaking concrete and remolding geometry in a physical model may be accomplished more quickly using a CFD model.

In general, CFD modeling is less expensive and can be performed more rapidly, and the results are more easily communicated to concerned parties. The authors' experience shows that CFD modeling of intakes typically is one-third to one-half the cost of physical modeling. For sediment modeling, the cost differential usually is higher, up to an order of magnitude. So why is physical modeling still performed? It comes down to trust. While CFD functionality has made enormous strides in the past 20 years, experience with this technology in the hydro industry does not match the more than 100 years of experience using physical modeling. Therefore, hydraulic engineers recommend physical modeling in scenarios where CFD modeling is not perceived as being properly validated, such as irregular or non-standard site-specific conditions; complex hydraulic conditions; or the use of a non-standard design to improve project performance, constructability, or economics.

For hydro projects, CFD modeling typically is used to ensure performance before going to the expense of actually building a project. The performance metrics addressed most often are powerhouse efficiency, operation and maintenance costs, safety, and environmental concerns (such as sedimentation). The following three case studies outline the uses of CFD and physical modeling methods to optimize all of the above concerns.

Recent modeling work shows that the trend is toward using a combination of CFD and physical models, rather than one to the exclusion of the other. Three case studies provide an understanding of recent usage of both physical and CFD modeling in the hydro industry and why developers continue to find value in physical model studies.

Case study: Smithland low-head hydro intake

American Municipal Power (AMP) is developing a 72-MW powerhouse at the U.S. Army Corps of Engineers' Smithland locks and dam on the Ohio River in Kentucky. The powerhouse will contain three bulb turbines.

The powerhouse location, on the left descending riverbank, was selected for foundation, constructability, and environmental considerations. Project engineer MWH Americas and turbine-generator supplier Voith Hydro developed the intake geometry to provide economical powerhouse construction as well as acceptable turbine performance. A 1,500-foot-long approach channel that makes a 60-degree converging bend delivers flow to the intake. This intake is short relative to those for similar projects and features a relatively large convergence angle from the trashrack to the bulb.

Hydraulic challenges that could arise are flow direction changes at the entrance from the river to the approach channel and at the bend immediately upstream of the powerhouse. As a result, flow approaching the intake separates from the inside (right) embankment and helicoidal flow is induced by the outside (left) embankment in the bend. The challenge is to provide approach channel geometry that uniformly distributes flow to the units, oriented longitudinally with the unit and with no significant vorticity. Design concerns also included conformance with Voith Hydro guidelines for intake performance, as well as balancing efficient energy generation with construction cost.

The design approach involved using a CFD model to develop and compare alternatives, validating the CFD model with results from a physical model, and final verification of the selected geometry using the physical model. The CFD model, developed using Fluent software, simulated 7,000 feet of the river, including the entire river channel and the dam. The 1:60 Froude scale physical model simulated 3,200 feet upstream of the intake and included a portion of the main river channel. Strict conformance of physical models with hydraulic similiarity criteria (such as Reynolds and Weber numbers) to precisely define prototype vorticity generally is not possible with models smaller than about 1:20 scale. However, models of this size generally are not practicable for hydro projects. Experience has shown that models of 1:40 to 1:60 scale are adequate to model low-level vorticity. Models of this scale are considered a practicable means of evaluating vorticity, if used conservatively. Approach flow conditions were defined by baffled head boxes at the upstream and right sides of the physical model and were checked against flow conditions simulated in a comprehensive 1:120 scale physical model.

Key intake performance parameters are velocity distribution uniformity, flow angle, and vortex strength. CFD models accurately simulate velocity distribution and angle, but detailed vortex simulation is not possible and results are subject to interpretation. Physical models accurately simulate velocities, flow angle, and vortex strength, as long as the scale is sufficiently large to avoid viscosity and surface tension effects. Velocity distributions computed in the CFD model at the intake bulkhead slot were compared with those measured in the physical model at the same location. Comparisons were made using graduated color contours (see Figure 1). The CFD model also provided the ability to visualize the velocity vectors in the intake between the trashrack and bulb. Agreement between the results was considered sufficient to justify use of the CFD model for alternative development.

This multi-frame time-lapse photo shows the effects on navigation of the addition of a hydro plant at the Smithland Locks and Dam. The red dye tracks show streamlines during operation of the hydro facility.

The velocity distributions for the initial design were close to the Voith Hydro guidelines. Deviations from the guidelines were primarily due to the short intake resulting in a skewed vertical velocity distribution. Higher velocities were measured near the top of the water column.

This velocity magnitudes measured for the proposed 72-MW Smithland plant intake show sufficient agreement between the computational fluid dynamics (at left) and physical (at right) models.  This agreement was deemed sufficient to allow the use of CJD to investigate alternative intake geometries.

Four alternative geometries were tested in the CFD model to improve the distribution and evaluate potential cost savings. The most promising of the alternatives resulted in a construction cost savings by reducing rock and overburden excavation as well as improved performance based on reduced head loss and improved uniformity of flow to the turbines. Flow visualization techniques provided by the CFD modeling improved understanding of the intake performance and aided alternative development.

The physical hydraulic model was modified to incorporate the selected alternative. Results of the physical and CFD modeling were compared. The physical model indicated an improvement in performance of the end units and a decline in performance of the middle unit, whereas the CFD model indicated improvements for all three units. The physical model indicated an overall improvement in performance, and the results of both CFD and physical modeling were similar.

While the CFD modeling served a valuable function in this study, it did not eliminate the need for physical modeling. The Smithland physical model demonstrated that vortices were not a concern at this site. Confidence in the CFD model would have been diminished were it not for the physical model's ability to demonstrate the lack of vortices.

Case study: Effects of hydro plant on navigation at Smithland

The Ohio River is a major thoroughfare of commercial navigation. Commodities are shipped in barge trains that may reach 1,100 feet in length and 110 feet in width. Lockages at the Smithland facility, about 60 miles upstream of the Ohio and Mississippi river confluence, average 21 per day.

The hydro plant described in the previous case study has the potential to affect both operation of the locks and dam and downstream flow conditions. During normal river conditions, the hydro plant will use the first 55,000 cubic feet per second (cfs) of water, which normally is directed through the spillway. The plant will deliver this flow back to the river through a 2,000-foot-long tailrace oriented at about 45 degrees to the river channel and directed toward the lower approach to the navigation locks on the opposite bank. Project effects on flow conditions warrant investigation of potential effects on commercial navigation.

The key mission of the Corps' Ohio River District is to provide safe and efficient passage for navigation traffic. As such, the district provides standards and participates in physical model studies to evaluate potential project effects while developing mitigation measures. The primary Corps concerns are:

– Effects on barge trains entering and leaving the locks;

– Changes in bed sediment movement affecting the navigation channel; and

– Effects on dam stability.

CFD models are capable of simulating project effects on flow distribution, but the interface with the moving barge is not sufficiently developed. Therefore, Alden constructed a comprehensive fixed-bed Froude scale model of the river reach potentially affected by the project. Scales ranging from 1:100 to 1:150 typically are used, and this model employed a 1:120 scale. The Corps maintains a fleet of towboats with these scales to simulate barge traffic. The Smithland model reach included 2 miles upstream of the dam and almost 3 miles downstream. The downstream reach included the confluence of the Cumberland River, several large islands, and a bend, all of which added complexity. The model was constructed to simulate the current baseline operation as well as future conditions, including the project as proposed. The model was calibrated/validated to recorded river velocity distributions obtained using an acoustic Doppler current profiler. The model also was calibrated to historic water surface profiles by modifying channel roughness in the flood zone until the model velocities were within about 10 percent of the measured prototype velocities. Corps personnel reviewed the calibration.

The Smithland locks and dam is in an area with a sand bed that has exhibited depositional characteristics. Periodic dredging of the navigational channel is required to maintain adequate draft for barge trains. The Corps is concerned that operation of the hydro project may change sedimentation patterns and increase dredging requirements. A comprehensive CFD model of the river reach downstream was used to evaluate project effects on river bed movement. The model included a 19-mile reach downstream of the locks and dam, all features of the locks and dam, and the hydro project facilities (including the tailrace). The model was used to simulate movement of the streambed for historic flow conditions both with and without the project. Results of CFD also were used to determine the flow distribution around the largest of the downstream islands as necessary to provide the boundary conditions for the physical model. The flow distribution predicted by the CFD model was validated by comparison to acoustic Doppler current profiler measurements obtained at two locations in the Ohio River.

The physical model was developed to simulate project effects over the entire range of operations: river flows from 7,000 to 250,000 cfs. The model has the capability to simulate higher river flows and floods up to about 700,000 cfs. Seven flow conditions, ranging from a single unit operating to the maximum navigable river flow, were selected for simulation. These conditions were simulated for two possible conditions of the Cumberland River as its flow is controlled by a hydro facility upstream. Operation of this facility releases additional flow into the river which, essentially, increases water level at a given discharge and affects the distribution of flow around Cumberland Island.

This overlay of the computational fluid dyanmics (CFD) and physical models of Canton Dam shows the service (at top) and emergency (at bottom) spillways.  Qualitative and quantitative agreement between the two models was excellent.

Corps representatives visited the physical model when validation was complete to simulate navigation under existing conditions. The Corps team included the lockmaster, lead hydraulic engineer from the Louisville District, and navigation experts from the Corps' Engineer Research and Development Center (ERDC). They simulated barge movement into and out of the upper and lower lock approaches and recorded the simulations using "stop action" digital photography. The lockmaster consulted with the ERDC barge operator regarding the appropriate navigation line and expected behavior. The ship tracks were then compiled to provide a visual record of the effects of river currents on the barges, and the operator documented his findings for comparison to later "with project" simulations. Alden prepared detailed records of river velocities and currents by using high-resolution digital imaging equipment to track the paths of drogues that float at the approximate draft of commercial barges. Software developed by Alden then was used to convert the digital images into velocity vector data files that were then plotted and overlaid onto drawings of the site.

The physical model then was modified to include the project facilities (powerhouse, tailrace, and approach channels), and work was undertaken to develop the optimum tailrace geometry. Tailrace geometry is important because of possible effects on navigation resulting from its orientation and velocities, which are higher than the existing river conditions. Personnel with Franklin G. DeFazio and project design engineer MWH Americas Inc. worked with Alden to develop a geometry that considers effects on energy generation (head loss), construction cost, and navigation conditions. Because the Corps was not present at these tests, the effects on navigation were estimated based on certain key flow conditions.

The physical model was used to develop a tailrace that minimized its length without affecting navigation. Corps personnel revisited the model after the tailrace design was developed and retested the same flow conditions with the plant operating. The tailrace design was modified slightly to improve navigation effects and project constructability.

Other Corps concerns also were evaluated using the physical model. Studies of "tracer" materials representing sediment were used to simulate patterns of bed sediment movement under baseline and "with project" conditions. It is not possible to precisely define sediment movements in a fixed-bed Froude scale model because sediment movement responds to bed shear stresses and requires similarity of shear forces on the particles along the bed. This calls for a simulation of particle sizes and density conforming to sediment scaling criteria. However, it often is acceptable to use fixed-bed models to qualitatively evaluate potential projects effects on sedimentation by comparing the movement of tracers in an existing condition model to "with project" conditions. This type of study generally is used to guide judgments as to whether there is the potential for significant effects (i.e., whether the project will direct sediment transport toward or away from a navigation channel).

Waves resulting from powerhouse load rejections also were simulated to determine the effects on barges in critical lock areas. In addition, the model was used to evaluate alternatives to mitigate potential effects to an endangered mussel species that inhabits an area near the project tailrace.

Case study: Spillway safety and design detail at Canton Dam

The Corps' Canton Dam is a 15,140-foot-long rolled, earth-filled dam with a maximum height of 68 feet above the riverbed. Construction of the dam, on the North Canadian River in western Oklahoma, was completed in 1948. The dam has one service spillway with 16 tainter gates and three outlet works sluice tunnels with a combined discharge capacity of 352,000 cfs at the maximum pool elevation. To safely pass the new probable maximum flood (PMF) inflow of 634,000 cfs, a new emergency spillway has been proposed.

The Corps chose a Fusegate system from Hydroplus Inc. for the new emergency spillway. Fusegates are freestanding concrete "buckets" set side-by-side on a broad-crested weir to form a watertight semi-labyrinth weir/barrier. Controlled flooding of a chamber underneath each Fusegate at predetermined water elevations creates a tipping moment, and the Fusegates individually tip to release the flood water. Water is conveyed to the Fusegate chambers through a large-diameter (10 to 12 feet) intake conduit leading to an adjacent intake tower. Inside the intake tower are multiple inlet wells, set at different elevations, that are connected to the Fusegates chambers through pipes embedded in the weir.

This overlay of the final emergency spill way shape on an aerial photo of the existing Canton Dam illustrates the scale and layout of the project.

Installing this system involves excavation of a new approach and return channel and installation of a 480-foot-wide broad-crested weir. This weir will be equipped with nine Fusegates, each 32 feet tall and 53.3 feet wide.

Alden conducted an integrated CFD and physical model study of the proposed spillway system to ensure that it can safely discharge the PMF. A CFD study was carried out first to evaluate modifications of the approach channel and reservoir geometries to achieve a cost-optimized configuration. The CFD model was used to compare approach flow patterns, resulting water surface elevations throughout the reservoir and spillways, and the flow splits between the service and emergency spillways. Based on the CFD result, a favorable design was selected, constructed, and tested in a large-scale physical model. The advantage of using this combined CFD and physical modeling approach is that each model can be used where it has its strength. It was faster and less expensive to investigate numerous approach channel geometries in the CFD model compared to constructing and testing all the geometries in the physical model. The physical model was used to validate the CFD results, obtain the new spillway rating curves, and make improvements to meet maximum water level criteria during PMF.

Eight cases representing a combination of six different geometries were analyzed. One CFD run was selected as a "promising design" for validation in the physical model. Froude scaling was used to construct a 1:54 undistorted physical model. The model boundaries were selected to represent the reservoir over a distance of about 2,500 feet upstream and 2,700 feet downstream of the service spillway. The model footprint was 110 feet by 70 feet, and the maximum model flow rate was about 30 cfs. The physical model inflow boundary flow conditions were adjusted to match the velocity profiles between the CFD and physical models (see Figure 2 on page 48).

Water surface elevations measured in the physical model closely matched the CFD model elevations. However, for more generally accepted water surface elevations and rating curves, the physical model must be used. Rating curves were obtained for each spillway separately and for combined spillway operation. The latter showed that the pool elevation for the PMF case still was unacceptably high.

Relying on data from the CFD and physical models, it was determined that an acceptable reservoir water surface elevation could be reached through several modifications in the physical model. These included removing a partially submerged berm and lowering the auxiliary spillway sill by 4 feet while also increasing Fusegate height by 2 feet. The Fusegate tipping sequence also was modified to lower the maximum reservoir PMF outflow. The maximum pool elevation for the PMF was thus reduced to an acceptable elevation. The physical model was then used to determine the optimal Fusegate tipping order, evaluate Fusegate evacuation through the return channel, select the intake conduit entrance location, and determine the intake well operating elevations.


The authors acknowledge the assistance of Howard Park and Randy McCollum of the U.S. Army Corps of Engineers, Engineering Research and Development Center; Russell Wyckhoff, Tulsa District of the Corps; Phil Meier, Assistant Vice President of Hydro Development at American Municipal Power; Christ Konstantellos, Lead Civil Engineer for MWH; Chris Miller of Alden; and Martin Wosnik of Alden and the University of New Hampshire. 

Timothy Sassaman, an engineer with Alden Research Laboratory Inc., was project manager for the model studies. Eugene Gemperline, PE, a consulting hydraulic engineer with Franklin G. DeFazio Inc., was lead hydraulic engineer. Kenneth Halstead, PE, is a hydraulic engineer with the Huntington District of the U.S. Army Corps of Engineers. Ken Lamkin, PE, is district hydropower coordinator with the Louisville District of the Corps. Hasan Kocahan, is a manager with Hydroplus Inc.. 

This article has been evaluated and edited in accordance with reviews conducted by two or more professionals who have relevant expertise. These peer reviewers judge manuscripts for technical accuracy, usefulness, and overall importance within the hydroelectric industry. 

More Hydro Review Current Issue Articles
More Hydro Review Archives Issue Articles

To access this Article, go to: