Civil Engineering Reference
In-Depth Information
by ignoring the simulation program's possible errors and approximations, and his
goal is to implement and then evaluate the effects of some external changes on the
model output.
The external methods are subdivided into local and global methods. In a local
method the effects of change of an individual parameter on the uncertainty of the
output of the model is evaluated, while in a global method the effects of change of
multiple parameters on the uncertainty of the model is assessed. The notable simi-
larity between the two approaches is that the correlation between the input and
output for both methods are usually considered to be linear, while the main differ-
ence between the two different approaches is that the input parameters in local
method are sampled one by one and the target is to understand the partial derivative
of the output in relationship with the input, but in a global method the input param-
eters are sampled all together at the same time and the target is to understand the
uncertainty of a specifi c input in relation to the overall outputs.
13.5.1
Local Methods
Local methods can be sub-categorized into differential sensitivity method and facto-
rial method. A differential sensitivity analysis method is usually used to evaluate the
effects of change of an individual input parameter on the uncertainty of the output.
In this method generally three sets of simulations (1) with original input value,
(2) with upper limit of input parameter, and (3) with lower limit of input parameter
is performed and the results from these three sets of simulations will be analyzed to
provide a better understanding of the effects of the individual parameter changes on
the output of the model simulations. This method is easy to perform and interpret
results, but the weakness of the method is that each input is assumed to be indepen-
dent from all the other inputs.
In the factorial method, during the simulations all the uncertain parameters alter
between either the upper and lower limits of input parameter or among upper limit,
lower limit and mean value of the uncertain parameters, and then the results will be
evaluated to provide a better understanding of the effects of the parameter changes
on the output of the model simulations.
13.5.2
Global Methods (Sampling-Based Methods)
Typically in global methods the uncertainty in input parameters is used to determine
the probability distribution of the output while all the variables are sampled simul-
taneously. In general the fundamental concept of all the sampling based methods is
to generate and investigate an overall depiction of relationship between uncertain
analysis inputs and uncertain analysis results. In another word each of the uncer-
tain analysis results are functions of the uncertain analysis inputs, and uncertainty in
