PERFORMANCE EVALUATION OF THREE

DAM FAILURE ANALYSIS MODELS

byAravind V. , Asst. Engineer, PWD, Govt. of Kerala

Muthiah Perumal, Professor, Deptt. of Hydrology, IIT Roorkee

N.N.Rai, Director, CWC, New Delhi

Third National Dam Safety Conference IIT Roorkee

18-19 February, 2017

ORGANISATION OF PRESENTATION

Introduction

Objectives

Models considered

Features of models

Considered case study

Results

Performance evaluation

Conclusions

Introduction

Necessities:

• For predetermination of the warning time and probablemaximum water level downstream of dam for variouslikely dam failure scenarios to plan evacuation measures

• To simulate the flood propagation of a recorded faileddam scenario

Approach followed for dam failure analysis Development or identification of the inflow hydrograph

to the reservoir at the time of failure.

Routing that hydrograph through the reservoir

Development of failure condition of the structure

Calculating the outflow hydrograph from the failedstructure

Modeling the propagation of the flood wave of the faileddam downstream to determine travel time, maximumwater level reached , inundated areas etc

Objectives

1. Reconstitute the flood wave resulting from the failure ofMachhu dam II which happened on 11th August 1979using the models NWS-DAMBRK(1990),MIKE11(2008) and HEC-RAS(2008)

2. Compare the results of each of these models with theobserved values and analyse the results.

3. Evaluate the performances of the above mentioned damfailure analysis models.

Models for analysing dam failures

The U.S.National Weather Service (NWS) Dam-BreakFlood Forecasting Model(DAMBRK) by Dr. D.L. Fread(1984)

MIKE 11 by Danish Hydraulic Institute (DHI).

HEC-RAS of U.S.A.C.E

Major Components of Analysis

1. Reservoir routinga) Storage Routing b) Dynamic Routing

2. Breach simulationa) Overtopping failure b) Piping failure

3. River routinga) One-dimensional b) two-dimensional

NWS-DAMBRK Model featuresReservoir routing

by Puls Method or Dynamic routing

Breach Description

a) Breach due to overtopping b) Breach due to piping

Simulating breach in the form of rectangular, triangular ortrapezoidal shapes.

Assumptions are to be made to approximate the actualbreach profile to correspond to any of these breach profiles.

By assuming equal area the breach is approximated to atrapezoidal form as shown:

River routing

One-dimensional routing using the St. Venant’s equationsbased on four point-implicit finite difference scheme

i. Conservation of mass (continuity) equation:

ii. Conservation of momentum equation:

Breach due to overtopping Breach due to piping

MIKE 11 MODEL FEATURES

Core of MIKE11 system is the HD module.

MIKE11 HD module is an implicit, finite difference modelwhich solves Saint Venant’s equations using 6-point Abbottfinite difference scheme.

HD module utilizes a space staggered grid consisting ofalternate h & Q points

During simulation, the St. Venant equations are solved numerically atthe grid points at specified time interval for the given B.C.

The i/p files used in dam failure analysis are:• Network file• Cross-section file• Boundary file• HD parameter file• Simulation file

Results of HD simulation consists of a time series of water levels anddischarges along the reach d/s of the failed dam.

Results can be viewed by MIKE View

HEC-RAS Model features

Dams are modeled using the Inline Structure editor.

Inline Structure editor allows the user to input anembankment, define overflow spillways and weirs, andgated openings.

Gate openings can be controlled with time series of gateopenings or using the elevation control gate operationfeatures in HEC-RAS.

Solves the St.Venant equations using the four-pointimplicit scheme.

Two c/s should be given before the dam.

The routing reach is hydraulically connected to thereservoir with the 1st c/s.

The 2nd c/s is required as a B.C. for the inline structure (thedam)

Considered case study Machhu Dam-II in Gujarat State failed on 11th August 1979

Relevant design aspects of Machhu Dam-II before the failureType of dam Masonry spillway with earth dam flanks on

either side

Length of the earthen dam on left 7689 ft (2343.61m)

Length of earthen dam on right 4588 ft (1398.42m)

Length of non-overflow masonry portion 272 ft (82.91m)

Spillway length 676 ft (206.0448m)

Shape of spillway Ogee

Crest of spillway RL 168 ft (51.21m)

Spillway design flood for avg. annual rainfall of558.8 mm

2,18,330 cusecs (6182.42 cumec )

Details of radial gates of the spillway 18 gates of 30 ft (9.144m) Long and 20 ft(6.096m) high

Low water level RL 155 ft (47.244m)

Dead storage 7926 ac.ft. (9776923.91m3)

Full reservoir level(FRL) 188 ft. (57.30m)

Gross storage 81520 ac.ft (100557006.9m3)

Flood cushion 1 ft. (0.3048m)

High Flood Level 189 ft. (57.61m)

Free Board 8 ft. (2.44m)

Top of the dam RL 197 ft. (60.05m)

The observed spillway flood which caused the failure was 13139cumecs due to 711.2 mm rainfall recorded in less than 24 hours. The water level has risen 1.5 ft. above the dam crest level whichleads to the overtopping failure at 1:30 pm.

3600 ft 1850 ft

Adopted final breach size

Adopted design inflow hydrograph into the reservoir of Macchu Dam-II(source: Jagdish Narain et al. (1981))

Adopted Manning’s n values and contraction-expansion values d/s

Results1. DAMBRK Model simulations

Simulated discharge hydrographs downstream of Machhu Dam-II using DAMBRK

2. MIKE 11 simulations

Simulated discharge hydrographs downstream of of Machhu Dam-II using MIKE 11

3. HEC-RAS simulations

Simulated discharge hydrographs downstream of Machhu Dam-II using HEC-RAS

Performances evaluationComparison of breach outflow

Comparison of time to peak stages at downstream of Machhu Dam-II

Comparison of simulated and observed peak stages at d/s reach

Comparison of computed and observed peak flood water levels at d/s reach due to dam failure

Conclusions On the basis of the simulated peak stage profiles with the observedvalues, the results from MIKE11 are much closer to the observed valuesalong the downstream channel than the other two models.

DAMBRK Programme shows high rate of attenuation, while theMIKE11 shows the least rate of attenuation and HEC-RAS showsmoderate rate of attenuation.

MIKE11 is more user friendly without involving trial and errorapproach in the use of spatial and temporal step sizes in comparisonwith the HEC-RAS usage.

In DAMBRK Programme the cross-sections at which dischargehydrographs can be displayed is limited to six.

Results can be represented in a more user friendly form such as by theuse of animation both in MIKE11 and in HEC-RAS models.