UNIT 10 RISK AND RELIABILITY

ANALYSIS

10.1 Introduction
Objectives

10.2 Methodology for Rislc and Reliability Analysis

10.3 Design Frequencies

10.5 Answers to SAQs

10.1 INTRODUCTION
Since plaillli~lgand design in hydrology are concerned witli fi~txmcvcnts which are
stochastic in nature, coilcepts of probability or I'requerlcy are necessary while dealing
with these aspects. The acceptable level of risk in water resources engineeriilg,
however, depends on econoinic and policy consideratioils. A ltydraudic structure can be
designed for the worst possible event, but the cost illvolvcd will be very high. So risk is
esthated, for a given frequ~ei~cy of a11 event (rain or flood); and bascd 011 acceptable
risk, liydr~llllicstn~cLuresare designed,
The designer of ally l~ydrologicstmclure is faced will1 the most i1nportm.t
question-what is the risk of its failure'?The price of failure of a dam is high due to loss
of lives and property; hence, the risk of failure of a darn ill~istbc ininin~ized.A study of
over 1600 dalus (Biswas and Chatlerjee, 1971) has sbow~lthe causes of failure of duns
as given in Table 10.1.

Table 10.1: Ciluses of Failure of D;lrns

S. No. Ci~usesof Frlilure of Dams Ln percent (YO)
1 Fo~u~dation
proble~lu 40%
2 I~ladequatespillway capacity 23 %

-
3 - Poor conslnlction 12%
4 , U11eve11settlemnent 10%
5 Hidl pore pressure 5%
6 Acts of war 3%
7 En~l>alllclnel~t
slips 2%
8 Defective materids 2%
9 Incorrect operation 2%
10 Ealthq~~akes 1%

Another study (more recent) o-fover 300 dun disaslers, however, shows that roughly
35% of the failures were due to inadequate spillway design (Biswns and Chatterjee,
1971). Inadequate spillway design @tisicallyinadequate le1lgCI1of the spiilway) is
usually the; result of inadequate design flood analysis; and tlzis is the direct concern of
the l~ydrologists.
Design floods for various hydraulic structures are eslimated either tlxougl~bequency
techniques, or a design flood is taken as the probable maximun~flood. The technique of
probable maximum flood, despite its name, is a totally deterministic concept and as
Statistical sand Pnobal~ility such has no risk associated wit11 it, as there is no proof of the existence of extreme
hlethotls in IIydrolog~
boundaries in lneteorological [actors which cause floods. Yevjevich (1968) states that
the concepts of maximum probable precipilation, inaxiinuln probable flood 'and other
similarly named i~naginalyevenls inay be considered as arbitrary, they being the
collcepts of expediel~cyonl!. Frequency analysis, on the other hand, accepts events of
any magnitude iis being possible though as Lhe inagnih~deincreases the probability 01
occurrence decreases (Kite, 1977).
The simplest procedure regarding frequency analysis to estiinate the design floods for a
given spillway is to select a return period and use either graphical teclmiques or a
rnatl~ematicaldistribution to derive the corresponding event magnitude
McCaig and Ericksolt (1959) state that in the past it has bcen a common practice to
design major dmls [or floods having theoretical rehlrn periods of upto 10,000 years.
The ASCE Hydraulics Divisioil Committee on Hydrometeorology (ASGE, 1973) has
snggested that the probable maximum flood is perhaps equivalent to a design period of
10,000 years. This elemei~taryprocedure takes no account of the increase of risk with
increasil~gproject life, and does not consider the ecoilol~licallyoptiinuiu design.

After studyjng tl~isunit, you should be able to:

r the concepts of risk analysis ai~dthose underlying Binomial and Poisson
distribntions, and
e use and significance of dirferent frequency le~lelsused in hydrologic design.

10.2 METHODOLOGY FOR RISK AND

RELIABILITY ANALYSIS
While designing any hydraulic stnlchire sonie of the questions which need addressing
to are:

i) vvllat is the need for arisk analysis'?

ii) vvllal should be the return period Tor which the given sh-tzcture should be
designed?
iii) how iliuch is the risk involved cvllen we design a stn~clure,having a design
lire of n years, for a flood of T-year rehim period?
iv) how ~nuchrisk is pennissible?
The return period for which a structure should be designed is estimated on tllc basis of
Ll~eacceptable risk. Risk is conceived as the probability of occurrence ofafloorl at lea,st
once during the successive years qf desigvl I@. Risk that is acceptable depends upon
the eco~lomicand policy considerations.
The general fonnula for the calc~~lalion or risk can be derived easily. If for a t h e
invarialt lxydrologic system the probability of occcurreiice of an event, x, greater than
the design event, xo, d ~ u i i ~agperiod of M years is P, then the probability of
i~on-occunei~ce of x, Q,is 1-P. If this design event has a reh~rnperiod of T years, 'and
its corresponding anilual probability of exceedance, p is given by:

Therefore, we can state t11ai the probability of its non-occurrence in airy one year can be
expressed as:
 Therefrom the priilciples of statistics the probability of non-occurrence 01this event in Risk and Reliability Analysis
n years is given by: .

Hence, the probability that x will occur at least once in these 17 years, i.e., the risk of
failure of the slri~ctme,X is:

R is the risk of failure of llle structure during n years,

T is the retnill period (chosen by tile collcenled autllority) for which the stnlcture
has to be designed, and
n is the design life of the stnlcture.

For rr hydraulic stnlch~rehavi~lga design lifc of 100 years, what will be the risk
itlvolvcd if it is designed for: i) 50-year return period flood, ii) I 000-year rctul.11
period flood'?
Sslu tion
i) Ill ilic first case n = 100-yea~s,T = 50-years and the risk (R) illvolved niay be
compnted by substituting the values of vl and Tin Equation (10.4)

ii) In the secol~dcase, n = 100 years, T = 1000 years and the lisk ii~volved'R' is:

Based on the level of risk that is acceptable, the rch~nlperiod lor which tlle structure
should be designed is decided upon. Table 10.2 gives relunl periods associated wit11
various degrees of risk and expected design life using Equation (10.4); and Figure 20.1
presents a graphical solution of Equation 10.4.

What return period must a highway engineer decide ~lponin the design of a critical
underpass drain if he is willing to accept: i) only 10% risk that flooding will occur
in the next. five years, ii) 20% risk that flooding will occur in l~cxt2 years?
Solution
111case i), the risk involved (R) = 0,10,
YI = 5 years,
T=?
 Statistical and Probability Tl~erefore,froin Equation (10.4), we have:
Metl~odsin Hytlrology

or,
i
0.10 = 1 - 1 --

7'- 48.1 years.

This mems that there are 10% chances that a 48.1-year flood will occur once or
inore tllan once it1 next five years.
Siniiiarly, in case ii), tlie risk u~volved
R = 0.20,
n = 2 years,

T =?
I

Therefore,

or, T = 9.47 years.

It shows that there are 20% cha~cesthat a flood correspondiilg to a reku11 period of
9.47 years will occiw the next 2 years. These values may also be read from the Table
10.2.

EXPECTED PROJECT LIFE, YEARS

Wgure 10.1 : TheoreticnlProbabilily of Failure, for given Project Lifc md Design Return period
(Kite, 1977)
Table 10.2: Return Periods Associated with Varions Degrees of Rislc and Expected Risk and Reliability Analysis
Design Life

"I'

10.2.1 Some Standard Probability Dist~.ibutions

1

Any sanlple of a llydrological variable call be analysed in 'two ways: tllrough

descriptive statistics, and inferential statistics. The fonner procedure analyses tile
histolical data to yield n distiibution of an event (data) Illat is kllowil as the observed
distsibuljoil (i.e., einpirical frequency distribution), and the11 it is approximated to one
of tile kllowil tlleoretical distributions with a view to enable evaluating the fhture
perfonnalce of structures. It is.obvious that tlle basic assumption it1 this approach is
tllat the future data sainples shall possess tile properties that are sane or siinilar to those
of the available historical sanlple-bul, this assumption lriay not be very t n ~ in e practice.
It may, however, be pointed out that any theoretical model (distribution) is never all
exact represei~tationof the give11natural process; it only is a possible description of the
process, that is useful in defining the obseived 'data.
Tllere are varicrus theoretical models available for use, sucll as, ~inomialdistribution
through Nonnal and Extreille Value distributions (some discrete and some continuous
frequency distribu~tions).
Tlle discrete probability distributions (Binomial and Poisson) are generally applied to
those random events in wllich the outcome is either an occurrence or ilon-occ~urence
i.e., there are two mutually exclusive events of a process. In this process it is assuiiled
that tlle successive events (tr'ials) are independent, and the probability of occurrence
(say, success) renlains consta~tfroin trial to trial,
If one is interested in knowiilg wllelher it is exactly once, twice, tl~riceor 'my given
ii~uinberof tinles that a given event (say, a flood) tis sly occur, the coiicepts regarding
Binomial distrib~ltioilor Poissoil distribution are used.
Binomial Distribution
The Binomial distribution, as tlientioiled above, is a discrete distribution a~dis based
011 Binomial theorein which states that probability of exactly x successes inn trials is
given by:

n!
x! (n - x ) !

p = is the probability of exceedmce (success),

q =probability of non-exceedance (failure),
x =number of exceedmce (successes), and
n =total uuinber of events.
Statistical and Probability The asswnptio~~s for Binomial distributioil are same as for Bernoulli trials. Tossing a
Methocls in Hydrology
coin or drawing a card from a pack are examples of Bernoulli trials which operates
under the following three conditions:
i) any trial call result in either success or failure, bue or false, rail1 or no rain,
ii) successive trials are hdependent, and
iii) probabilities are stable.
Binoinial distribntion is valid ~lnderthe above three conditions. It is clear that the
distribution applies to populations that havc oilly hvo discrete but conlple~nei~tary
events-c.g., rainy and dry days. Therefore, q is the probability of the coinplemeiltary
event (say, EL dry day).

If a d i l l is haviilg project life of 50-years h e n what is the probability that a flood

with return period of 100 years will occur: i) once, ii) twice during the life of the
dam.
Solution
i) n = 50 years,
T = 100 years,
The probability ofexceedance y is:

Here,

TIus means Illat there are 30.6% chances that 100-year retunl period flood will.
occur once during tile project life.
ii) n = 50 yews,
T = 100 years, ,
 This ineans that tllere are 7.5G% chances that the 100-year retunl period flood will Risk :tncl Reliability Analysis
occur two tllnes duling the project life.
Poisson Distribution
The terms of a Binomial expansion are a little incoilveilientto colilpute in any large
number. If n is large (>30) andp is sinall (c0.1) wit11 their product ny tellding to a
constant, A, then Biiloinial distribution tends to Poisson distribution, namely:

where, h = np.
It is not necessary in practice tllatp + O and n -+ w; it is, however, sufficient i f p is
relatively small and n relatively large f i r the Poisso~ldist.ribution to be reasonably .
applicable. Therefore, it call be applied to any extre~lievalue problem when the
occurrence of a11 event of interest has a probability p illat is proportioilal to tlle period
of observation, n, such as fulding the probability of
i) drougbts in a given span of time,
ii) number of rainy days at a given location,
iii) a rare flood, like 1 in 100 years or so, and
iv) a reservoir being empty in ally one year out of many years.
The conditions for this approximation are as listed below:

i) the number of events is discrete;

ii) two events cailuot coincide,
iii) the m e a l nuinber of events i11a unit time is constant, and
iv) events are illdependent.
The above Example 10.3 can be solved with Poisson distribution also. Its solutiion
using tlle Poissoil distribution is given below:
For this case, h= 17p =-50/100= 0.5

0.52- e-0-S
ii) ~ ( 2 ) = 21 = 0.0758 = 7.58%

It is observed thal the rcsulls obtained by Binomial distribution md Poisson distribution

are almost same.

10.3 DESIGN FREQUENCIES

A designer of water resources stn~cturesis very much coilcenled with tile return period
for which tlle stnlcture should be designed. 111 India, the present practice Bureau of
Indian Standars (BIS) Code is to specify the design flood for various slmctures in terms
of its frequency or return period. The design criteria for so111e of the hydraulic
structures is discussed in this section.
i) Criterin for Fixing Spillwny Capacity
The l~ydrologicdesign criteria for fixing spillway capacity as prevalent in India are
mentio~ledin IS 11223-1985, Guidelines for fixing Spillway Capacity, According to
these guidelines various inflow design floods that need to be considered for various
functions of spillways are:

.il
I
St;rtistical ancl Probability Inflow design flood for the safety of the dams
Metl~o~ls in Ifytlrolog'
a)
It is the flood for which the perfonllailce of Llle dm1 should be saCe against overtopping,
I stl~~ctural failure and with respect to its energy dissipation arrallgemeilts (if provided
1
for a lower flood) that should r~mctionreasonably well.

b) Intlo~vdesign h o d for efficient operation of energy dissipation works.

This flood could be lower than tlte flood for safety of dam and for this the dissipation
arra~gements,are expected to perfonn most efficiently.
c) Illflow design flood for checking the extent of upstreal11 submergence.

d) Inflow design flood for tlle extent of downstream damage in the valley.
The criteria for the classificsltioilof dams is based 011 the size of the dain and the
llydraulic head (MWL-average flood level on the dawvlzstrean). The classification Tor
the dam is determined by the greater of tlzese two panlleters:

Classiiication Gross Stomgc Hy drzmlic Hea 11

(in ~ ~ u l l i ocubic
n meters) (in metcrs)
Snlall Between 0.5 and 10 Betweell 7.5 and 12
Intcnnediatc Between 10 t ~ n d60 Retwecn 12 and 30
Large Greater than 60 Cireater than 30

The inflow design flood for the safety of the dan vis-a-vis, its size is stunniarised as
follows:

Size (as determined above) Inflow Design Flood for Safety of Dnin
Sr~lall ' 100-year flood
Standarc1Project Flood (SPF)
Large Probable Maximum Flood (PW)

Floods 01 larger or s~llallermngnitudcs may be used for tlle safety analysis if the hazard
involved is high or low. The relevant parameters to be considered in judging the hazard
in addition to tlle size would be:

a) diststilce to and location of dle humau habitations on the downstreanl side after
co~lsideril~g the likely future developments.

b) maximum hydra~~lic capacity 'of the downstrean cl~al~ilel

at a level at which
catastrophic dzunage is not expected.
For more iillpoitallt projects dsun break stutdies may be done as an aid to Ll~ejudgment
ill deciding tvbetl~erPMF needs to be used. Where the studies or judgment indicate iul
iillminent danger to present or fuhwe 111unm settleme~~ts, tlte PMF shonld be used. Any
departwe fro111 the general criteria as given above, on account of' larger or smaller
hazard, should be clearly brought out and recorded.
Probable Maximum Flood
Probable ina~imumflood is the flood caused by probable inaxirn~u~l precipitation,
Probable inaxilllum flood is geilerally obtained by using unit hydrograph and rainfall
estiillates of PMP. T l ~ probable
e maxi~nnmstonn is defuled as h e most severe storm
coilsidered reasonably possible to occur. The ASCE Hydraulics Divisio~lCommittee on
Hydroineteorology (1 973) has s~lggestedthat the probable i n a x i ~ n ~flood
~ m is perl~aps
eqt~ivalentto a desigll return period of 10,000 years.
Standilrd Project FlBods
SPF is the flqod cansed by stwdard project stornl wllich is gel~erallyobtained fioin a
survey o f severe storms 111 the gelleral viscinity oftlle drainage basin or severe storms
experiei~cedjn i~~eteorologically
similar areas.
 -

ii) Criteria for Design Flood Estimation for Barrages Risk and Rcli:~bilityAnalysis

For barrages, tlle CWC (Central Water Commission) criteria (1968) are applicable.
Diversioil dans or weirs and barrages have usually sinall storage capacities, and the
risk of loss of life and property do.vvnstreamwould rarely bc enhanccd by failure of the
stnlcture. However, allart froin the loss 01property and stl-~~ctures by its failure, this
would bring about disn~ptionof inigatioil and cominunications that are dependent on
the sdcty of the barrage. In coilsideratioil of thesc risks involved, tl~eCWC criteria
were refonllulated Lo include floods of frecli~ency50 to 100 ycars for design purposes.
For bairages, it requires the use of a 100-year return period flood or staildad pro-jecl
flood whichever is higher.
iii) Criteriil fo~.;DesignFlood Estimation for Weirs (Ungated Hetld\vlirl<s)
111ilie case of sniall'resei-r~oirswhere the relcasc of stored water ~ L I to
C the failure of the
dun would not appreciably enhance the ilood hazard dowl~stream,the spillway
capacity inay be dcsigned for a design flood of specil'ied frcquci~cy,say 50 to 1011 years
as recommended by tlle Ceiltral Water Commission.
iv) C1.ite1-iafor Dcsign Flood Estimatioa of Road and Railway Bridges
For road bridges, thc Indian Road Congress, [RC: 5- 1970, Section-1 (Gcneral Fcalures
or Design) applies. Accordirig to this: tlle design discharge for whicll tlle walenvay of a
bridge is to be dcsigncd shall be either the mzlr;imwu flood observcd for a pcriod ofnot
less than 50 ycars; or shall be the discharge fioiii ai~otlicrrccogniscd method applicable
for that area; or shall be the discharge found by the area velocity method; or shall be
the discl~argefound by unit-llydrograph method-that is il shall be the maximuin-
discharge fixed by ilie judgme~ztofthe engineers respo~wiblefor the design by
comparing all llle results oblaiiled by these ~lletllods.
For railway bridges, a 5.0-year flood is to bc uscd for sillallcr bridges cniryi~lgrailways
of lesser iillportailcc like minor lincs and brmlch lincs. I11 Ihc case 01' larger bridges, i.e.,
those carryiilg illail1 lines and vely ii~lporla~it
rail lines, a 100-year rcR~riipcriod ilood is
lo be adopted as per i l ~ railxvvg~
e codes (Indian Railway Standards - 1963)

v) Criteria for Design Flood Estimation for Cross Dr:lin;tgc Structures on

Irrigatioll Netw o 1-l<s
The BIS Code of practice for the design of cross'dmii~a~c works [IS : 7784 (Part - I)
19751 recoilu~lendsthat tllc design (of watci~vay)in such cases may be based on 10 to
25-year freq~leiencyflood with increased ai'llux. Howevcr, tllc rou~ldatioilsand
free-board etc., should be checked to be safe Tor the increased nEnu~xmd velocitics duc
to a 50-year or a 100-year rehm pcriod llood.
For very Iarge cross drainage works, dallagc to which is likely to affect Ihe canal
supplies over a loilg period the design sl~ouldbc based on li~axi~nulll probable flood. It
is quite probable that n flood olhigher magnit~~de than the design flood lnay pass
througll the stnichlre posing great dangcr to the stability of rot~ildsltiolland the
stnlct~re.Adequate returii period to take cire o r this uilprecedcnted and unrorsee~l
natt~reof flood illte~lsitiesin case of iinportailt sl~ucluresis assumed to afl'ord or an
adequate margin of safety is envisaged in tllc estinlation of dcsign discharge. For this
pnrpose, the design $ischarge niay be increased by Lhe pcrcelltages give11 below Tor
obtainillg the foundatioil and free-board dcsign.

Catchment A r c a (in square Idlomcters) Incrcase in Design Dischnrgc

upto 500 30% to 25% dccrensillg with imrease in sreu

500 to 5000 25% to 20% decreasing with increuse in area

5000 to 25000 20% to 10% decrensing with illcrease in area

above 25000 upto 10%)

 I Statistical an11P~.obability
I As per Central Water Commission criteria, waterways for canal aqueducts should be
Methodsin Hydrology
provided to pass a 50-100 year retun1 period flood, but their foundations and
~1 free-boards s h o ~ ~be
l d checked for a flood of not less than 100-yezir return period.
The Govenlment of Gujarat llas adopted a still severer criteria for cross drainage works
of Sardar Sarovar N m l a d a Canal: which are given below:
--

Catchment Area Design Floucl to be Adopted

(in square miles)
For Design for Checldng
0 to 10 100 year ilood 100 year flood + 30%
10 to 50 -do - -do -
50 to 200 - do - PMF
200 and above - do - PMF
(or SPF )

vi) Design Criteria for Flood Control Schemes

The following broad criteria are recom~llet~ded
al~dadopted in the country:

Predomii~antlyagricultural 25-year retunl period flood on slllall

tributaries and 50-year flood on nlajor
rivers
Town protection works 100 - year return period flood
Important ir~ddustrialcomplexes, assets and 100 -year return period flood
lines of coilllnunications

According to Gmga Flood Control Conlmission, subject to availability o l observed

hydrological data, the design HFL may be fixed on the basis of flood frequency
analysis. l i no
~ case, the design HFL shou~ldbe lower tl~anthe tnaxinlum on record. For
small rivers carrying discl~argeupto 3000 cumec, the design I-IFL shall correspolld to
25 years rett~nlperiod flood. For the river carrying peak flood above 3000 culnec, the
design HFL shall correspond lo 50 years reh1n1 period. Howevcr, if the embailkments
concen~edare to protect big township, industrial area or olher places of strategic
importance tlle design HFL shall generally correspond to 100-year rehun period flood.
The Rashtriya Barh Ayog reco~nme~lds that benefit-cost criterion should be adopted for
optimum design of stn~cturesvis-a-vis, flood control measures. But, since the relevant
data for suzch an analysis may not be available, the Ayog recoi~~mends t l ~ efollowiilg
guidelines: i) for predominailtly agricultural areas adopt a 25-year flood frequ~ncy(in
special cases, where the damage potential justifies, higher values may be adopted); and
ii) for town protection works, important industrial complexes, etc, adopt a 100-year
flood frequency (for large cities like Delhi, the lnaxitnuln observed flood, or even the
mixin1u.m probable flood should be considered for adoption),
Each site is unique i n its local canditions, and for the evaluation of causes, and effects.
Therefore, the above mentioned nonns, may be taken as the general guidelines, but the
hydrologist, and, the designer would have the discretion to vary the norms, and the
criteria in special cases on a c c o ~ uof
~ tassessable and acceptable local conditions, These
considerations should be recorded, and, have the acceptance of the competent authority
(Sharma, 1991).
SAQ 1
a) If the probability of occurrence of an event is 35'%1, what is the probability of
its non-occurrence'?
b) If any design event has a return period of 100 years, what is its probability of
occurrence'?
 ibvs
'sai1q31~qsloauo:, p o o -103
~ p u ~ s ~ p o no?lu8!~~;r
~au
110 STTOMa8eu!e~pssolr, 'sde~\l!el prrs sptroi ' s n a ~ t'saSaneq 'sdtr~11!dsj o u8rsap
3!1nwp6q alp i o j '.:,la ' s p o o ~u$!sap
j Sn~sooy:,~ opale1rmuoj
j rtaaq antry a3p3eidjo
sap03 put! sap;rtlD .suo~lt~q;r.qs!p uoss!od pup pyutouya su rpns 'srto!lnqyqsyp k p ~ q ~ q o ~ d
p n p t ~ e ~arrros
s jo asrl arp sasscduroona s!s6~tlrm ri1?[!qo![a~ pue ys!~j o k 8 o ~ o p o y ) a ~
.@adold prm aj!I jo ssol
~rtanbasuo:,o) aorralaja1 I~!M sallqolys pa)qaJ alpjo amlyejjo ys!~pa)r,adxa all$qlIM
s))rraAa3~2o~olpAq
pa3cj SF iarrqlsap arlL 'anqerr n! uropuul ale spool^ '~I~!JU!~.I) astle3aq
L1?1!q~qoldjos~da3uo3ayl aqoArn rrB!sap puc Butmme1d Suuaau19rta s a ~ m o s aia)vM
i