Verification and Validation of HEC-RAS 5.1

Transcription

1 Verification and Validation of HEC-RAS 5.1 Gary Brunner 1, P.E., D. WRE, M.ASCE Dr. Alex Sanchez 1 Dr. Tom Molls 2 Dr. David Parr 3 1. USACE Hydrologic Engineering Center, Davis, CA 2. David Ford Consulting Engineers, Inc., Sacramento, CA 3. Civil, Environmental and Architectural Engineering, University of Kansas. US Army Corps of Engineers 1

2 HEC-RAS 5.1 Verification and Validation Research Study HEC is performing a comprehensive Verification and Validation study for HEC-RAS 5.1. This will cover: 1D Steady Flow 1D Unsteady Flow 2D Unsteady Flow The following types of data sets are being used for this research work: Analytical and Text book data sets Laboratory experiments Field data (real world flood events with observations) 2

3 Current 2D Analyses Performed Analytical and Text Book data sets: 1. Chow Steady Flow Backwater profiles 2. Subcritical flow over a variable sloping bed 3. Flow transitions over a Bump 4. Flood Propagation over a Flat and Frictionless plane 5. Dam Break on a Flat and Frictionless Bed 6. Sloshing in a Rectangular Basin 7. Long-wave Run-up on a Planar Slope 8. Surface Runoff on a Plane 3

4 Current 2D Analyses Performed Laboratory Test Cases: 1. Flow in a Compound Channel 2. Flow in a Rectangular channel with a Sudden Expansion 3. Two-dimensional Surface Runoff 4. Partial Dam Break on a flat plane 5. Rectangular channel with a 180 Degree Bend 6. Sudden Dam Break in a sloping flume 7. Converging Channel (Sub to supercritical flow) 8. Dam Break in a channel with a 180 degree bend 9. Flow around a Spur Dike 10. Flow through a Bridge opening 4

5 Current 2D Analyses Performed Field Test Cases: 1. Malpasset Dam Break 2. New Madrid Floodway, May 2001 Flood 3. Hopefully more??? 5

6 What I am presenting today: The following are 2D Verification/Validation data sets: Sloshing in a Rectangular Basin Analytical Test Channel with a sudden expansion Lab Data Flow Through a Bridge opening Lab Data New Madrid Floodway, May 2011 Flood Event Field Test 6

7 Sloshing in a Rectangular Basin The sloshing test case is useful to verify the temporal scheme implementation and analyzing the numerical dissipation as a function of time step and numerical scheme. Leandro, J., Chen, A.S., and Schumann, A A 2D parallel Diffusive Wave model for floodplain inundation with variable time step (P-DWave). Journal of Hydrology, [In Press] Hunter, N.M., Horritt, M.S., Bates, P.D., Wilson, M.D., Werner, M.G.F., An adaptive time step solution for rasterbased storage cell modelling of floodplain inundation. Advances in Water Resources. 28, BUILDING 7 STRONG

8 Problem and Data Description 10,000 m long and 300 m wide closed rectangular channel The depth of water is 10m deep. The analytical solution was derived with the following assumptions: No Friction No advection No turbulent diffusion Eliminating the velocity from both equations leads to the classical wave equation: 8

9 Starting Wave Conditions 9

10 Model setup Model Parameter Water depth Wave amplitude Wave length Value 10 m 1 cm 10 km Gravity m/s 2 Time steps Simulation duration Grid resolution 0.5, 5 s 10 hrs 100 x 100m Implicit Weighting Factors 0.6, 1.0 Manning s roughness coefficient s/m 1/3 10

11 Results and Discussion Water Surface Profiles WSE profile goodness-of-fit statistics Time 00:34 02:54 7:29 09:49 RMSE (10-4 m) NRMSE (%) MAE (10-2 m) NMAE (%) R

12 Results and Discussion Theta = 0.6 and DT = 0.5, 5.0s 12

13 Results and Discussion DT = 0.5s and Theta = 0.6,

14 Results and Discussion In general, the computed water levels agree very well with the analytical solution as demonstrated by the goodness-offit statistics. The differences between computed analytical water levels increase with time due to numerical dissipation (possibly some friction loss are still there and loss due to acceleration terms). There is a small asymmetry in the water level which gets more pronounced with time, this is due to the diffusion and differences in travel time, which grows over the period. The Numerical dissipation is more sensitive to the computational time step than Theta. 14

15 Sudden expansion Channel with a sudden expansion creating an eddy zone Xie, B.L. (1996). Experiment on flow in a sudden-expanded channel, Technical report, Wuhan Univ., China. Reported in: Wu et. al. (2004). Comparison of five depth-averaged 2-D Turbulence models for river flows, Archives of Hydro-Engineering and Env. Mech., 51(2), BUILDING 15 STRONG

16 Test facility Bottom width b = 0.6 m to 1.2 m Side slopes z = 0.0 Bed slope S = Roughness n =0.013 for the bottom and for the sides Flow rate Q = cms and cms Down Stream Boundary Condition, WS = m and 0.1 m L 4.6 m 18 m 16

17 Model Setup Mesh cell size: dx = 0.05 m Computation time step: dt = 0.1 s, Cr = Vdt/dx 1 n = (concrete) D T = 0.55, eddy viscosity coefficient (0.1 < D T < 5, from RAS 2D User s Manual) S 0 = 0 BC: Q u = and cms; h d = 0.1 m Full shallow water equations Q u = and cms h d = 0.1 m 17

18 Results (Eddy Zone for Q = ) D T = 0.55, eddy viscosity coefficient L RAS matches experimental reattachment length L RAS = L exp 4.6 m L RAS 4.6 m Vmag (m/s)

19 Results (Velocity Profiles Q= cms) X=0 m X=1 m X=2 m X=0 X=1 X=2 X=3 X=4 X=5 L exp 4.6 m X=3 m X=4 m X=5 m 19

20 Results (Velocity Profiles Q= cms) X=0 m X=1 m X=2 m X=0 X=1 X=2 X=3 X=4 X=5 L exp 4.7 m X=3 m X=4 m X=5 m 20

21 Sensitivity Test (Vary D T, Eddy Viscosity Coefficient) Reattachment length is dependent on D T Increasing D T reduces L RAS D T =0.0 L RAS 5.3 m D T =0.55 L RAS 4.6 m Vmag (m/s) D T = L RAS 4.0 m 21

22 Results Summary Computed eddy zone reattachment length matches experimental length (with D T = 0.55) for both flow rates. Computed transverse velocity profiles closely match experimental profiles. This is an interesting test case because it requires modeling turbulence. 22

23 Flow Through a Bridge Opening The purpose of the test case is to validate HEC-RAS 2D Version for simulating flow through a bridge opening. The laboratory flume had left and right rectangular overbanks and a trapezoidal main channel. Abutments and bridge piers were located in the middle of the test reach. HEC-RAS 2D model (Model) results are compared to measured laboratory piezometer readings. The Model was developed assuming a 20-scale undistorted representation of the laboratory flume. This Test Case is from the research work of Dr. David A. Parr, Professor; Evan Deal, Research Assistant; and Bryan C. Young, Associate Professor, all are in the Civil, Environmental and Architectural Engineering Department at the University of Kansas. BUILDING 23 STRONG

24 Test facility The laboratory flume had left and right rectangular overbanks and a trapezoidal main channel. Abutments and bridge piers were located in the middle of the test reach. Length = 24 ft long Test region Bed slope S = 0.0 Horizontal Roughness n = Constant Head Tank Upstream, Overflow Flap Weir downstream 24

25 Test facility Piezometer Locations 25

26 Test facility Bridge Cross Sections 26

27 Model Setup Terrain Data 27

28 Model Setup Mesh Setup (20,578 cells) 28

29 Model Setup Nine Experiments were performed Flows Q = 3620, 4400, and 5600 cfs Three Tailwater elevations for each flow rate Mesh cell size: 2.0 ft general, and 1.0 ft along breaklines Computation time step: dt = 0.5 s, Cr = Vdt/dx 1 Full shallow water equations Test No. 6 was excluded from the numerical experiments, as it was not a stable condition. The flow was jetting out of the downstream side of the bridge opening and also oscillating from left to right downstream of the bridge. 29

30 Results 22 Run 1, 2 and 3, HR max WSE ( t = 0.5 sec) 24 Run 7, 8, 9, HR max WSE ( t = 0.5 sec) WSE max (ft) WSE max (ft) Station (ft) Run 1, 0.5 s Run 2, 0.5 s Run 3, 0.5 s Lab Run 1 Lab Run 2 Lab Run 3 Bridge Channel Bank Station (ft) Run 7, 0.5 s Run 8, 0.5 s Run 9, 0.5 s Lab Run 7 Lab Run 8 Lab Run 9 Bridge Channel Bank 22 Run 4 and 5, HR max WSE ( t = 0.5 sec) 24 Run 7, 8, 9, HR max WSE ( t = 2 or 4 sec) WSE max (ft) WSE max (ft) Station (ft) Run 4, 0.5 s Run 5, 0.5 s Lab Run 4 Lab Run 5 Bridge Channel Bank Station (ft) Run 7, 2 s Run 8, 2 s Run 9, 4 s Lab Run 7 Lab Run 8 Lab Run 9 Bridge Channel Bank 30

31 Results Summary The results for Runs 1, 2, 3, 4, and 5 are in excellent agreement with the lab data throughout the piezometer reach. The HEC-RAS 2D profiles for Runs 7, 8 and 9 were high for t = 0.5 seconds as shown. Consequently, other time steps were tested to determine the best fit t-values of 2, 2 and 4 seconds for Runs 7, 8 and 9, respectively. The results of this study are very encouraging. The HEC-RAS 2D model did exceptionally well in matching the laboratory data. 31

32 New Madrid Floodway Levee Breach The performance of HEC-RAS in simulating a real life levee breach and interior flooding, is demonstrated with the New Madrid Floodway data set. During the May 2011 flood in the Mississippi and Ohio rivers, the New Madrid floodway was activated by blowing up the levee in three locations, in order to reduce flooding along the Mississippi and Ohio rivers. Before the levee was activated, the U.S. Geological Survey (USGS) went out and placed 38 pressure transducers within the floodway, in order to measure the water levels during the event. The levee system was activated by the US Army Corps of Engineers on May 2, The resulting floodwater (measured water surface elevations) inside of the floodway are used to compare against the computed values from HEC-RAS. This test case is useful for comparing HEC-RAS 2D results to real world data. BUILDING 32 STRONG

33 Location Map of New Madrid Floodway and Breaches 33

34 Problem and Data Description The New Madrid Floodway is located just below the confluence of the Ohio and Mississippi rivers. This floodway is surrounded by a levee system, which is designed to be activated during major floods. Approximately 9400 feet of the upper levee was activated on May 2, 2011 just after 10:00 p.m. Later, two lower sections of the levee were also activated to allow water to drain out more efficiently. Terrain data for the floodway was obtained with LIDAR (5 ft grid res) Surveyed cross sections for the Mississippi and Ohio River systems. Levee locations and elevation data was obtained from the National Levee Database (NLD) Flow data for upstream boundaries was obtained at USGS gages A rating Curve was used for the Downstream Boundary Breach Data was obtained from the U.S. Army Corps of Engineers 34

35 Model Schematic with 38 USGS Measurement Locations shown as Red Dots 35

36 Model Setup Flood way is 40 miles long and 10 miles wide Grid Resolution used was 500 x 500 ft cells Manning s n values were set within the floodway by using a Land Cover data set. Parameter Value Time step 2 minutes Governing Equations Shallow Water Equations Implicit Weighting Factor 1 Manning s roughness 1D rivers main channel overbank areas Upper Levee Breach width 9400 ft. Middle Levee Breach width 690 ft. Lower Levee Breach width 4100 ft. 36

37 Results for all 38 Locations Location Computed W.S. (ft) Observed W.S. (ft) Difference (ft) H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H Ave ABS Error Summary of Results: Max Difference = 0.68 ft Min Difference = 0.0 ft Ave Abs Error = 0.29 ft

38 Computed vs Observed Water Surface at locations H1 and H7 38

39 Computed vs Observed Water Surface at locations H8 and H10 39

40 Computed vs Observed Water Surface at locations H12 and H17 40

41 Computed vs Observed Water Surface at locations H20 and H24 41

42 Computed vs Observed Water Surface at locations H26 and H33 42

43 Questions? 43