State of the Art Two-Dimensional Hydraulic Modeling

ToolsModel Benchmark

Testing

Presented by:

Murari Paudel, Cody Milligan, Soledad Roman, Phillip Ryan

September 2015

Presentation Overview

• Study Purpose

• Methodology and Limitations

• Test Cases

• Assessment Results

• Conclusions

Study Purpose

• Benchmark TUFLOW (2013), MIKE21 (2014) and HECRAS 5.0 (March 2015)

• Test the models for a range of basic hydraulic conditions and model performance measures:1. Hydraulic losses: Does the model scheme

account for losses accurately?2. Grid convergence: Does the model produce

similar result for a range of cell sizes?3. Simulation run time: How efficient is the

model’s solution scheme?

Methodology

• Create 3 simple test cases1. Channel bend2. Constriction3. Flat bed (flood propagation)

• Run simulations for a range of flow conditions and cell sizes– Number of scenarios = 109

• Assess model results against measured data

1

2

3

Test Case – Bend Loss

• Aim: verify models estimate losses resulting from an abrupt bend accurately

• Validate against laboratory experiment (Malone 2008)

Test Case – Bend Loss

• Empirical equations from laboratory results

Bend – Model Configuration

• Scaled to match dimensions of laboratory test

• 0, 45 and 90 degree scenarios. 15ft grid resolution.

• Free overfall conditions downstream boundary

InflowBoundary

Downstream Boundary

Bend – Model Configuration

• Scaled to match dimensions of laboratory test

• 0, 45 and 90 degree scenarios. 5 flow scenarios. 15ft grid resolution.

• Downstream boundary free overfall conditions

Output Location P1

Output Location P2

Result: HECRAS5.0 (Full Momentum)

Result: MIKE21

Result: TUFLOW

Bend Test Result Summary

• All models produced reasonable results– MIKE21 and HECRAS 5.0 resulted in slightly greater

head loss due to bend– TUFLOW resulted in slightly smaller loss at 90◦ bend

and slightly more at 45◦

• Runtime EfficiencyHECRAS 5.0 (FM) = 2 minutes 36 second (8 Processors)MIKE 21 = 1 minute 23 seconds (8 Processors)TUFLOW = 2 minutes 13 seconds (1 Processor)

Test Case – Constriction

• Aim: verify models estimate losses resulting from a constriction (eg. bridge abutments)

• Aim: assess grid convergence in model results

• Validated against Federal Highways Administration Equations (Bradley, 1978)

Constriction – Model Configuration

• 40% opening width in a uniform rectangular channel

• 10ft and 30ft grid resolution. 3 flow scenarios.

• Free overfall conditions downstream boundary

InflowBoundary

Downstream Boundary

120ft

90ft

90ft

Result: HECRAS5.0 (Diffusive Wave)

Result: HECRAS5.0 (Full Momentum)

Result: MIKE21

10ft Resolution, 3000cfs

Result: TUFLOW

10ft Resolution, 3000cfs

Constriction Test Result Summary

• Grid size convergence – HECRAS 5.0 Full momentum, Mike21 and TUFLOW produced

reasonable results• HECRAS 5.0 and Mike21 produced slightly higher loss while TUFLOW

produced slightly smaller loss

– HECRAS 5.0 Diffusion wave showed inconsistency

• Runtime Efficiency (10ft resolution)HECRAS 5.0 (DW) = 3 min 4 sec (8 Processors)HECRAS 5.0 (FM) = 4 min 34 sec (8 Processors)MIKE 21 = 1min 29 seconds (8 Processors)TUFLOW = 2min 3 seconds (1 Processor)

Test Case – Speed of flood propagation

• Based on UK EPA benchmark test

• Aim: “test is to assess the package’s ability to simulate the celerity of propagation of a flood wave and predict transient velocities and depths”.

• Aim: Compare model simulation run time

• The model scenario size has been increased from UK version to make a better estimate of the model simulation time (50,000 cells)

Flood propagation – Model Configuration

• Flat bed

• 20m grid resolution

• 20m3/s inflow through a 20m opening

• 24hr simulation period

InflowBoundary

8000m

4000m

P2P1

Result: Water Level

Result: Velocity

Result: Wave Propagation

Result: Simulation Time

• Simulation TimeModel Simulation Time 20m

HECRAS 5.0 27 minutes 53 sec

MIKE21 7 minutes 3 seconds

TUFLOW 56 seconds

Conclusions• Model Results?

– Channel Bend Loss – All three software produce relatively similar results• HECRAS 5.0 and MIKE21 marginally overestimate the head loss

• TUFLOW is marginally underestimating

– Channel Constriction – Consistent results across all the three software• HECRAS 5.0 and MIKE21 produced head loss slightly higher

• TUFLOW is marginally underestimating

– Wave Propagation• All the three software show consistent results, timing, depth and velocity

• Significant difference in the model run times

– HEC-RAS 5.0 - Diffusive wave… Use with caution

• Model Simulation Times?– MIKE21 and TUFLOW similar. HECRAS 5.0 relatively slower– Parallel processing vs parallel simulations?

• Other factors?

Thank You

Murari Paudel, PhD, CFMWood Rodgers Inc.

mpaudel@woodrodgers.com

(916) 341-7760